Field

Values

What a position is worth — numbers, and the things that are not numbers, and how to find the simplest one.

Four things a position can be. Every position falls into one of four outcome classes, and only three of them correspond to a comparison with zero. The fourth — first player wins — is a position confused with zero, neither greater, smaller nor equal, and it is where the subject departs from arithmetic.

Who moves last

The player who cannot move loses. That single convention generates the whole theory — and it produces four outcomes rather than three, because a position can be confused with zero rather than greater, smaller or equal to it.

The simplest number in between. A game whose options are numbers is worth the simplest number strictly between them — and simplest means born earliest, so integers come before halves and halves before quarters. It is not the midpoint, and the difference is the whole content of the rule.

The simplicity rule

When both players' options are numbers, the position is worth the simplest number strictly between them. Not the midpoint, not the average, and the difference between "simplest" and "middle" is the entire content of the rule.

The same game, written twice. A position as it arises and the same position reduced. Left would never move to −1 when 0 is available, so that option is dominated and can go. The two games are equal — checked, not assumed — and the second is the canonical form.

Canonical form

Two positions are worth the same when neither player can tell them apart inside any larger game. Deciding that could be an infinite search. Instead there is a normal form — delete what nobody would play, bypass what backfires — and equality becomes a comparison of two small trees.

Infinitesimals

Some positions are positive — Left wins them whoever moves first — and smaller than every positive number, including a millionth and a millionth of that. They are the values that decide close games, and the smallest of them is a single move's worth of nothing.

$The numbers, by the day they are born. Zero on the first day, ±1 on the second, and thereafter the simplest number in each remaining gap. Every number reachable in finitely many days is a fraction with a power of two underneath, and every such fraction appears — which is a strange thing for a construction with no arithmetic in it to produce.$

The day a number is born

Start with a position in which neither player can move, apply one rule, and the numbers appear — but only the fractions with a power of two underneath, and only in a particular order. That order is what "simplest" means.

Why nobody moves in the number. A hot position added to a number. Left wins the sum whoever moves — but only by moving in the fight. Spending the move on the number instead hands the position back as a first-player win, with Right to move, which throws the win away. The theorem says this is always so, and here it is happening.

Numbers avoid numbers

In a position with a number in it and anything else, the number is never the right move. That is a theorem rather than a heuristic, and it is the closest this subject comes to advice a player can carry into a real game.

A switch, its mean and its temperature. Positions of the form {a | b} with a above b: both players want to move there, so neither is settled. The bar spans the two options, the marked point is the mean the position is worth once the fighting is over, and the temperature is half the gap — which is exactly what moving first is worth.

Worth nothing, and worth fighting for

A switch is a position both players want to move in. Its average value can be zero while the difference between getting there first and second is enormous, and that gap is a second number every position carries.

The same game, written twice. A position as it arises and the same position reduced. Three of the options are dominated — a sibling is at least as good for the player who owns them — so they can go. The two games are equal — checked, not assumed — and the second is the canonical form.

Two hundred and fifty-six ways to write twenty-two things

Every game whose options come from the four born on day one — there are 256 of them, and between them they carry 22 values. The reduction that collapses one to the other has choices in it at every step, and uniqueness is the claim that none of the choices matters.

Tiny and miny: infinitesimals with a scale. Positions that are greater than zero and smaller than every positive number, and which are nevertheless strictly ordered among themselves — the larger the subscript, the smaller the value. Being smaller than everything positive is not one size of thing; it is a whole scale, and up sits above all of it.

Tiny, miny, and the sizes below every size

An empty two-by-four Domineering board is worth less than nothing and more than every negative number. It is not up, not down and not a fraction — it is a miny, and the minies come in sizes, strictly ordered among themselves below a floor no number reaches.

Clobber: every value smaller than every number. Blue and red stones on a small board. A move takes one of your own stones onto an orthogonally adjacent enemy stone, which is removed. Because adjacency is symmetric, a player has a move exactly when the opponent does — so no position can ever be worth a whole move to anybody, and every value that comes out is an infinitesimal.

The class where nobody runs out first

Three stones in a row — blue, blue, red — and the position is worth exactly up. Clobber cannot produce anything else, because adjacency is symmetric — a player has a move precisely when the opponent does, and a game with that shape can never be worth a whole move to anybody.

$The days this site can compute, and the ones it cannot. Zero on the first day, ±1 on the second, and thereafter the simplest number in every remaining gap — the construction run by the game recursion, which produces only fractions with a power of two underneath however long it goes on. Below it, three objects the same recursion reaches when the stopping rule is removed, each written with its option set and the exact reason this site's machinery cannot hold it. They are named rather than drawn, which is the honest half of a figure-first collection.$

The recursion this site cannot run

Remove the stopping condition from the construction and it reaches ω, its reciprocal, and one third — none of which this site's evaluator can represent, because it interns a position from a finite list of options. The figure draws what it computes and names what it cannot, which is where the boundary belongs.

7 positions of the same value, and how long each of them lasts. Nim positions whose heap sizes all nim-sum to zero. As games they are the same object: each is worth zero, each is a loss for the player to move, and each may be substituted for any other inside any sum without changing a single outcome. The bars are how many moves each one takes, from the shortest legal play to the longest. The value determines everything about who wins and nothing at all about when.

What a value leaves out

A value settles who wins, by how much, and what happens in every sum the position appears in. It says nothing about when. Seven positions here are worth exactly zero and interchangeable everywhere, and they run from two moves long to eighteen.

The thermograph of {5 | 1}. Temperature runs up the page and value across it. Each wall is where a player is willing to move once a tax of that much is charged per move; above the temperature at which they meet, neither wants to move and the position is worth its mean value. The height of the meeting point is what is at stake. The two marks on the base line are the stops — what each player gets by moving first and playing the fight out with no tax charged at all.

Where the fight stops

Keep taking the biggest thing on offer until somebody is left facing a number, and the number reached is a stop. Every position has two of them, one for each player moving first, and between them they say how a fight ends — except for the part of it no number can reach.

What a move is worth to the player making it

The gain from a move is the option minus the position it was played from — and that is a game rather than a number, so two moves can be incomparable instead of one of them being best. Temperature is what happens when the largest of those games is asked for a single number.

What a deeper position does to the shape. Thermographs side by side, two of them, with temperature running up each panel and value across it: {{6 | 2} | {1 | −3}}, with a bend where an option's own fight cools out; {4 | −1}, straight-walled. A wall that runs straight has nothing changing hands below the meeting point; a bend is an option's own fight cooling out at a lower temperature than this position's, and it is where a decision passes from one player to the other. The two marks on each base line are the stops — what each player gets by moving first with no tax charged.

The switch a player is imagining

Every account of a hot position ends up as "worth about m, and worth t to move in", which is the switch {m+t | m−t}. For a plain fight that summary is the position exactly. For a fight with anything behind it the leftover is not a rounding error — on one position here it is a whole second fight of temperature two.

How old a form is, and how old its value is. Every one of the 256 forms born by day two, placed by the depth it is written at and by the birthday of the value it carries. Nothing sits above the diagonal, because a form cannot be younger than the value in it; the diagonal holds the forms written at exactly their value's birthday, and everything below it is a position written older than it needs to be. The count in each cell was obtained by canonicalising all 256 forms and measuring both depths.

How old a value is

A form's depth bounds the birthday of the value inside it, and reducing to canonical form attains the bound — for all 22 values born by day two, with no exception. Twenty-four of the 256 forms are older than what they are worth. The same reduction that makes the bound tight is what puts day three within reach: 98 option sets a side instead of four million, 9,604 forms, 1,474 values, a quarter of a second.

When a switch is not a switch

A position {a | b} with numbers on both sides is a fight only while a is above b. Sweeping the boundary with a fixed at 2 and b climbing from −2 to 3 turns up three regimes rather than the two the definition suggests: eight fights whose temperature is exactly half the gap, one position at a = b that is 2∗ and is not a number, and beyond that numbers chosen by the simplicity rule — which on 8 of 12 sampled cases is not the midpoint.

The fight never runs backwards

Left's stop is never below Right's — in every one of 1,780 distinct values, computed twice by two independently written routes, with nothing that disagreed anywhere. The inequality is what makes a mean value well defined and a fight a fight; and where it collapses to equality, 433 of the 460 cold positions turn out not to be numbers at all.

Options handed to Left in 1 | −1. A position, and one candidate option after another added to it. Where the gift is one the player would never take the value does not move at all; where it is one they would, it does. The last column is the value of the enlarged form, computed by the same recursion as the original.

An option nobody would take

Every reduction of a form deletes. The gift horse principle adds: a move may be handed to a player for nothing, provided it is one they would never choose. Over all 484 additions to the values born by day two, 283 leave the value exactly where it was and the 201 that move it are precisely the ones the condition forbids — with the boundary at *not better*, which is a weaker demand than *worse*.

The 22 values born by day two, and the order they form. Each value sits above everything it is greater than, joined to what it covers. The order has 36 covering relations and is nine levels deep, and 52 of its 253 pairs are incomparable — and it is still a lattice: every pair has a least upper bound and a greatest lower bound among the same 22 values. Two values are marked, together with their join and their meet.

The simplest game above both

Values sit in a partial order, and a partial order is entitled to be ragged: two things with no least thing above them. The 22 values born by day two are not ragged at all. Every one of their 253 pairs has a least upper bound and a greatest lower bound among the same 22, and the order is distributive on all 10,648 triples — so it is a lattice, and the join of zero and star is one half.

What each move is worth to the player making it. For each position: every incentive, whether they are all strictly negative, whether the position is a number, and its temperature. The middle two columns are two different computations of the same fact.

Nobody wants to move here

A position is a number exactly when every move loses ground for the player making it. The test never mentions numbers, it disagrees with the ordinary one on none of the 1,474 values born by day three — and the reason a position fails it is not that somebody wants to move. It is that somebody cannot afford to wait.

The reduction that puts options back

Canonical form is presented as simplification, and half of it is. Deleting a dominated option takes one away. Bypassing a reversible one substitutes the answer's whole option list, so it can leave the form wider than it started — and 60 of 32,428 forms end up with a canonical form wider than they are.

What deleting is worth on its own. The reduction split into its two halves and each measured. Deleting a dominated option removes exactly one option and can do nothing else; bypassing a reversible one substitutes an option list and can widen the form. The counts say how much of the reduction the monotone half accounts for.

The reduction that always shrinks

Canonical form is two reductions and they are not the same kind of operation. Deleting a dominated option removes one option and can do nothing else; bypassing a reversible one substitutes a whole option list. Over the 256 forms born by day two, deleting alone finishes 225 of them and accounts for 480 of the 520 options that come off — and the 31 it cannot finish are almost all the ones with a star in them.

The values the construction hands down, and the values games produce. The two lists counted against each other. The construction produces 1,474 values by day three; the eleven thousand positions swept here produce 1,193, and only 116 of those are on the construction's list. A value's birthday and a value's reachability have nothing to do with each other.

The values nobody's game produces

The construction hands down 1,474 values by day three. Seventeen rulesets on this site, swept to eleven thousand positions, produce 1,193 — and only 116 of those are on the construction's list. Two of the twenty-two values born by day two are produced by no position of any game here, and 1,077 of the values that are produced are born later than day three. A value's birthday and a value's reachability have almost nothing to do with each other.

How much a list of options can lose

Deleting a dominated option is the reduction with no surprises, and how many options it takes is decided by the shape of the order rather than by the values in it: the survivors are the maximal elements, and the count is the length of the list less the number of them. The essay separating the two reductions closed by predicting that the longest chain would give the number. It is a lower bound, exact on 3,859 of the 7,315 four-option lists and wrong on the rest.

The numbers each position is confused with. Each row is a position. The bar runs from its right stop to its left stop; the filled part is the set of numbers the position is genuinely confused with, computed one comparison at a time. The two coincide except at the ends, and a position whose stops meet is confused with nothing at all even when it is not a number.

The numbers it is confused with

A position is confused with a number when neither is at least as good as the other, and the set of such numbers is an interval. It is exactly the open interval between the two stops: over 36,850 comparisons the rule is wrong nowhere it speaks, and the 2,596 comparisons it declines are precisely the ones at an endpoint, where the position and the number differ by an infinitesimal.

How many options a value needs. The canonical form is the smallest form of its value, so the number of options it carries is a property of the value. Three days of the construction, with the widths that occur and the widest value of each.

How wide a form can get

Bypassing a reversible option replaces it with a whole option list, so a form grows in the middle of its own reduction. Whether it can come out wider than it went in is the question that leaves standing, and over 64,515 forms built from day-two options the answer is no, not once — the growth is real, it is transient, and the widest canonical form reached is exactly as wide as the widest form that reaches it.

What a value costs to write down

The canonical form is the smallest form of its value, and it is smallest in the one currency the reduction happens to spend: options. Counted in symbols it is nothing of the kind — the widest value born by day three is not the longest, the longest has six options rather than seven, and every canonical form on the day except the seven integers writes some position out twice.

Which end of the interval is open

The confusion interval is open at both ends, and the two ends are not the same kind of open. At its own left stop a position can be below the number, confused with it or equal to it, and — over 2,948 comparisons — above it exactly thirty-three times, every one of them a value whose two stops are the same number and whose left end is therefore also its right one.

How old a sum is. Every unordered pair of the twenty-two values born by day two, with nought dropped because adding it settles nothing — 231 sums. The birthday of each sum was read off its own canonical form and compared with the sum of the two parts' birthdays, which is the bound. The bound holds everywhere and is attained 163 times.

The birthday of a sum

Two values born by days m and n have a sum born by day m + n at the latest, which is the bound that stops a board made of many small parts from being unboundedly complicated. Over 231 pairs of day-two values the bound holds every time and is exact 163 times — and every pair it misses by three days or more has a sum that is a number or a nimber, so the slack is not noise but a measure of how much cancelled.

The identity that would join the order to the addition. Every pair of the twenty-two values born by day two, asked whether the join plus the meet equals the sum. It holds on all 201 comparable pairs, where the join is the larger and the meet the smaller and it cannot do otherwise, and on none of the 52 incomparable ones.

Where the order and the sum disagree

Day two is a lattice, and day two is a group, and it is not a lattice-ordered group. The one identity that would join the two structures — the join plus the meet equals the pair — holds on exactly the 201 pairs where it cannot fail and on none of the other 52, and the errors split thirteen high, thirteen low and twenty-six confused.

What a day of canonical forms costs, written out and written once. Three costs for the values born by each of the first three days: every node written every time it occurs, every distinct subposition of a single form, and every distinct subposition of any form of the day. The last is one node per value, and the gap between the first and the last widens as the construction goes on.

The same position, written once

Writing out the canonical forms of day three takes 24,940 nodes. Naming each distinct subposition once inside each form takes 10,102, and naming each distinct subposition once across the whole day takes exactly 1,474 — one per value, because nothing appears inside a canonical form that is not itself a value of the day.

The switch formulas, off the hypothesis they were stated for. Values born by day three with exactly one option a side, split by whether both options are numbers. On the twenty-one that satisfy the textbook hypothesis the midpoint and half-gap formulas are exact; on the 146 that do not, the same formulas read off the two stops instead hold about three quarters of the time.

A fight with no midpoint

The mean of {a | b} is the midpoint and the temperature is half the gap — on the twenty-one values of day three where a and b are numbers. One hundred and forty-six others have the same shape and not the hypothesis, and the repair that suggests itself, reading the two stops instead of the two options, holds on about three quarters of them and no more.

Fifty-two errors, put to four instruments. The fifty-two discrepancies the lattice identity leaves on day two, counted by what distinguishes them. As values no two are the same; as pairs of stops there are seven; as means three and as temperatures three. Not one of them is a number, and only three are values born by day two.

Fifty-two errors and seven sizes

Day two is a lattice and a group and not a lattice-ordered group, and the fifty-two incomparable pairs it fails on leave fifty-two different error terms. Measured rather than listed, the fifty-two collapse: seven pairs of stops, three means, three temperatures, and a rule that predicts the temperature from the pair on forty-four of them.

How old a value is, against how big a board it takes to show it. For each birthday, the range of sizes of the smallest position exhibiting a value of that age. The bars do not march rightwards: values born on the last day of the sweep are shown by six-piece positions, and values born on the fourth need up to fifteen.

The cheapest way to show a value

Eleven thousand positions from fifteen rulesets reach 1,193 values, and for each of them there is a smallest board that shows it. Set against the birthday the two measures agree hardly at all — until the numbers are taken out, at which point they agree rather well, and the whole apparent independence turns out to be a fact about integers.

Which description of a surviving option is right. The two candidate readings of what the reduction keeps, scored over every Domineering option list on six boards. Taking the most room is right on under half the lists, which is what a description with no content scores on lists this short. Leaving the opponent fewest replies is right on nine in ten.

Which option the reduction keeps

Domination deletes an option when another is at least as good, so what survives is the top of an order. On a board that order is made of moves, and two descriptions of the surviving move suggest themselves. Over 1,586 Domineering option lists one of them is right 47% of the time and the other 90%, and the one that wins is not the one a player would guess.

The birthday is a floor

The rung below measured a correlation of 0.73 between a value's birthday and the size of its cheapest exhibit, and asked which values are dearer than the birthday suggests. The relation is not a trend. Over all 728 non-number values the exhibit is never smaller than the birthday and is exactly the birthday on 476 of them, and the excess on the other 252 belongs to the game rather than to the value — the ruleset accounts for 40 per cent of its variance.

The bend decides it. Whether the stop reading gives the mean and temperature, against whether either wall bends below the meeting point. Both off-diagonal cells are empty on all 138 values.

The bend is the condition

The rung below offered a description of the class its stop reading is exact on — neither wall bends below the meeting point — and a route to proving it: that a bend happens precisely when some option is neither a number nor an infinitesimal. The first is exact on all 138 values, both directions, no exception. The second is half right: every bent value has such an option and 49 unbent ones do too. And the eight apparent exceptions to the first turn out to be a bookkeeping convention.

The value does not count the good moves. How many of a Domineering region's placements are best ones, against what the region is worth. Half the values have two regions that disagree about the count, and 52 disagree over whether there is a choice at all.

How many moves are worth making

A value answers who wins and by how much, and the anchor below names the quantities it discards. This is the first of them counted. Over 1,034 Domineering regions and 125 values, 63 values have two regions disagreeing about how many placements are worth making and 52 disagree over whether there is any choice at all — and the count of good moves stays near one and a half however large the region gets.

The margin the count needs. Every pair of Left options sorted by how many more replies one leaves the opponent than the other. At a margin of three the option leaving fewer replies is never the worse one.

The margin a count needs

Leaving the opponent fewest replies names only surviving options nine times in ten, which leaves the question of what a bound stated in that count would have to be weakened to. It is a margin. Over 57,879 pairs of Domineering options, the one leaving the opponent fewer replies is the worse of the two 1,052 times at a margin of one and 72 times at a margin of two — and at a margin of three, never.

Four ways to count a reply. The plain count of the opponent's replies against three weightings of it, each scored on the same pairs of Domineering options. Every weighting has a larger threshold than the plain count and gets more pairs wrong.

The weight that blunts the count

The rung below found that counting the opponent's replies gets the direction of a comparison right once the gap reaches three, and proposed a repair: weigh each reply by whether it leaves the opponent anything. Weighing it makes the count worse. The threshold goes from three to four, the failures from 1,124 to 1,320, and all seventy-two of the pairs the repair was written for come through it unchanged.

The error is half the follow-up. Each bent value's true temperature, the temperature the stop reading gives it, the difference, and half the temperature of its hot option. The last two columns agree on every value.

Half a follow-up out

The rung below settled which values the stop reading is wrong about — the ones whose walls bend — and left the size of the error unmeasured. It is not bounded by anything readable off the diagram; it equals something readable off the diagram. On all thirty-two, the mean and the temperature are each out by exactly half the follow-up's temperature, and the temperature is always read too low.

The trend, running backwards. The excess grouped by the width of the value's canonical form. Values with wide option lists are exhibited nearer their birthdays than narrow ones, which is the opposite of what the proposal predicted.

Wider costs less

The rung below found the cheapest exhibit of a value never smaller than its birthday, exactly equal on two thirds, and the ruleset explaining 40 per cent of the rest. The variable it proposed for the remainder was the width of the form. Width and excess correlate at −0.39: the wider the value, the closer to its birthday it is exhibited, and inside a ruleset the relation cannot even agree on a sign.

What a value costs a player. The number of positions a winning strategy has to tell apart, against the number of values among them. The value compresses 4,269 positions into 128 numbers and leaves 3,308 choices to be remembered.

What a strategy has to remember

A value answers who wins and by how much, and it settles neither how many moves achieve it nor whether the best one is unique. Counted over every position reachable inside the catalogue of regions, the gap has a size: 4,269 positions carry 128 values between them, and a player who wants to win rather than to predict has to store 3,308 choices — twenty-six entries for every number the theory supplies.

Every failure is on one board. The eight Domineering boards of the mobility census with the number of failing pairs on each. Seven of them contribute none; every failure at a margin of two is on the largest board, at two depths, and sixteen positions up to symmetry.

The threshold was a fact about the census

Two rungs failed to account for the seventy-two pairs where a mobility count gets the direction of a comparison wrong, and the third looks at them one at a time. They are not a class of shapes. All seventy-two are on the largest board in the census, at two depths, and sixteen positions up to symmetry — and one board larger the count fails at a margin of three, which the ladder has been quoting as the point at which it never does.

The correction on eleven times the pool. The stop reading and the corrected stop reading scored against the true temperature over every non-number value born by day three. The correction was measured on thirty-two values and holds on three hundred and forty-eight.

A second level of stops

The rung below found the stop reading's error to be exactly half the follow-up's temperature and asked whether the correction survives a wider pool, survives two bends, and can be stated without a thermograph. It survives eleven times the pool, missing two values in 1,459. It needs no thermograph — the follow-up's temperature is half its own stop gap. And it does not survive two bends, because day three contains no value with two of them.

The excess is not a flat fee. The excess fitted against the birthday inside each ruleset with enough values to fit a line. A fee would have a slope of nought; every slope here but one is negative, so the excess is largest on the values born earliest.

The entry fee was the cap

Two rungs measured how much bigger a position has to be than the value it exhibits, and attributed what was left to the ruleset — Toads and Frogs paying 2.25 squares on everything, green Hackenbush paying nothing. Neither number is a property of the rules. Inside every ruleset the excess falls as the birthday rises, because the sweep's size cap censors exactly the values that would pay most — and three squares past the cap, Toads and Frogs exhibits values born later than the strip is long.

The list saturates at three. Rules added greedily, each chosen to answer the most decisions given the ones already on the list. Three rules answer 94.5 per cent, and the fourth and fifth answer not one more.

Three rules and a tie-break

An exhaustive table of what a Domineering strategy has to remember is 3,308 lines. Three rules applied in order answer 94.5 per cent of it — leave the opponent fewest replies, then keep the region whole, then take whichever placement comes first — and the fourth and fifth rules answer not one more. The residue is 181 decisions in which every rule scores the candidates the same and one of them is worse.

The wall as an envelope. A thermograph with each option's contribution to its wall drawn over it, built from that option's two stops alone. The wall is the envelope of those contributions and it bends where the envelope has a corner.

The bend is in the stops

The rung below reduced the whole stop reading to one question — does this wall bend? — and asked whether that could be answered from the options' stops instead of from a diagram. It can, in four lines, and it gives more than the bend: on all 1,459 non-number values born by day three the options' stops determine the entire thermograph. One day deeper it breaks, and every failure is a value with a bent-walled option.

The rate was the alphabet

The rung below asked for a quantity a size cap cannot censor and proposed the rate: how many new values a ruleset produces per extra square. The rate is honest and it measures the notation — every ruleset grows at close to the number of symbols its positions are written in, and the seven span less than a factor of two. What separates them is the yield, which spans a hundred and nineteen.

Thirteen sweeps, four thresholds. The mobility rule's failures on every board and depth the sweep can afford, with the threshold each one gives. The thresholds take four different values and no ordering of the boards produces them.

A threshold is a detection limit

The rung below had two points — a margin of three at fifteen squares, four at eighteen — and asked whether the mobility rule's threshold grows with the board. Eleven more sweeps say no property of a board orders the thresholds, that the same board at two depths gives two of them, and that a tenth of the sweep which produced the four reports three instead. What does move, on every board measured twice, is the depth.

Two failures, not one. The decisions a Domineering strategy has to store, split by what the two content rules do: answer them, name a worse placement, or leave two candidates standing.

Seventy-two of them were not silence

The rung below said its 181 unanswered decisions were all the rules falling silent and asked whether the position's value picks the placement once the geometry cannot. Seventy-two of the 181 are the rules speaking and being wrong, which is a different failure. On the 109 that really are silence, a rule chosen per value answers more than half — and the star class the rung below singled out is settled outright by leaving the younger position.

How many levels, and how often. Every value in both pools by the number of levels of the recursion its thermograph needs before the stops suffice.

A bend that never reaches the surface

How many levels of the recursion a thermograph needs before its stops suffice is a number attached to a position, and the rung below conjectured it was the depth of the deepest bend in the tree. It is not: on 124 values a bend one level down costs nothing at all. What the number counts is the longest unbroken chain of bends running down from the top, exact on 2,400 of 2,403.

The mirror was the floor

Toppling Dominoes' share of genuinely new values had fallen from one to a half over eight sizes, and the rung below could not tell a floor from a slow fall. Four more sizes settle it: the distance above a half halves every two sizes. And the half is not a shortage of values but a symmetry — a row played from the other end is the same game, and the ruleset is as injective as that allows.

One board, all the way down. The mobility rule's failure rate on a three by five board at every depth, with the threshold each depth gives.

A heuristic that becomes a theorem

The mobility rule's failure rate had been measured at two depths on each of five boards and found to fall. Swept at every depth it does not merely fall — it accelerates, and it reaches exactly nought before the endgame. From four to eight empty squares onwards the rule has no exceptions at all, which turns a rule of thumb into a guarantee for the last few moves.

Wrong by one, or by nothing. The seventy-two decisions the rules get wrong, by how many replies the named placement misses a best one.

The price of taking the maximum

The seventy-two decisions where a Domineering strategy's rules name the wrong placement are never wrong by more than one reply, and a third of them are the second rule's fault rather than the mobility count's. The repair that follows — keep every placement within one reply of the best — retains a best placement every time and costs thirty decisions for every one it saves.

The theorem a proof would have needed. The mobility rule's failures on decomposed positions against connected ones, across every board in the depth sweep.

The easy case was not the reason

The rung below found the mobility rule reaching a failure rate of exactly nought near the endgame and named what a proof would need: that a decomposed board's comparable options are ordered by reply count. That statement is false on all five boards, at margins up to two — and split positions go exact two squares of depth before whole ones, so decomposition is the easy case rather than the cause.

Two of three are subadditive. The birthday, the option count and the written length, each tested for subadditivity under the disjunctive sum.

Two measures bounded, and one not

A sum is born no later than its parts' birthdays together, and it has no more options than they have between them — a bound nobody had checked, and it is attained. What runs away is the length of the written form: 27 pairs of 231 exceed it, the worst by 29 characters, on a sum with exactly as many options as it was entitled to.

A factor, and it grows. The ratio between reducing and deciding, by the number of options the form carries.

A factor, and not an overhead

Deciding who wins a form searches the form's own tree. Reducing it to canonical form searches a difference game for every comparison, and a difference game is a sum. Over 256 forms the reduction expands 5.46 times as many positions — and the ratio runs from 0.58 at one option to 9.80 at eight.

The value predicts on one residue only. Value-indexed rule selection on the silent decisions and on the seventy-two.

Close calls nothing resolves

The same value panel that settles 118 of the 202 silent decisions settles four of the seventy-two the rules get wrong. Every rule that helps at all must replace connectivity rather than follow it, and the cheapest one breaks twenty-six decisions for every one it saves.

Same group, three different yields. The rulesets with a trivial symmetry group, which the conjecture predicts must all have a yield of one.

Three groups, and three yields

The conjecture was that each ruleset's yield tends to the reciprocal of its symmetry group's order. Three rulesets have a trivial group and predicted yields of one; they measure 1.000, 0.531 and 0.204. And every colliding value in Push and Shove — all 175 of them — has two rows no symmetry relates.

The chain, scored. The chain reading of the level count against both pools.

The bend above the top

The chain reading gets three values in 2,403 wrong because it counts bends that the diagram never reaches. Counting only the bends below the position's own temperature fixes all three and breaks none — the first exact reading on this ladder, and it needs one comparison rather than the envelope the rung below expected.

The same game, written twice. A position as it arises and the same position reduced. Left would never move to 0 when 2 is available, so that option is dominated and can go. The two games are equal — checked, not assumed — and the second is the canonical form.

A reduction that reads a graph

The two reductions are defined as deletions from an option list, and the shared form has no option lists — a node is reached from several parents at once. Both restate as rewritings at a node, the rewriting is confluent, and its fixed point is the canonical form. What does not carry over is the sharing: four fifths of the shared nodes need a different answer under different parents.

The order one day out. Whether the values born by day three still form a lattice. Twice as many pairs are incomparable as at day two, and every incomparable pair still has a least upper bound and a greatest lower bound — so the order becomes more tangled without becoming ragged.

One of four questions

Three rungs of this ladder rest on sweeps of day two — 22 values, 253 pairs. Day three is 1,474 values and over a million pairs, and only one of the four questions can be asked of it. The order can: twice as many pairs are incomparable and every one of 1,606 sampled still has a least upper bound and a greatest lower bound, none of them a value day two already had. The other three compare sums of day-three values, which are born on day six, and sixty of those exhausted an eight-gigabyte heap.

All essays