Every essay — page 28
Sums and comparison
Real games break into independent parts. Adding them up is the whole method, and comparing them is how it is checked.
The follower does the reversing
The gift-horse theorem for the ordinal sum needs two cases, and the second — the added option is reversible — was counted and not described. Recorded move by move, the reversing answer is always Right's move inside the follower: on all 410 escapes under five followers, and on every one of the 2,628 gift horses under every follower that gives Right a move at all. The case split is by follower, not by horse.
The split slips one day deeper
The reversal case of the gift-horse theorem was described in one line — the follower's own move reverses every gift horse, whenever the follower has one — and tested only where it was found. In the mirror it holds exactly, with 1 and −1 trading places. One day deeper it fails: under ↑ and ½, three gift horses on built day-four bases are not reversed by the follower's move. All three are dominated, so the theorem stands; the clean split by follower does not.
Twenty draws and a second recipe
Day four's comparability was reported as 60.6 per cent against day three's 59.7, from one built sample and one calibration. Built twenty times with each of two recipes whose biases differ by six points, and calibrated against all 1,474 day-three values rather than a quarter of them, the corrected figure spreads over twelve points from seed to seed and the two recipes agree within one standard error. The floor survives; the decimal was one draw.
The same parts, different wholes
Equality is worth having because equal parts make equal wholes: a game may be replaced by an equal one inside any sum and any larger position. The relation that 184 contexts define — no context tells the two apart — is exact on day three and merges a third of a sample of day four. Put 567 of its merged pairs through negation, addition and forming options, and it survives two of the three. Adding the same game to both separates them 1,527 times.
What a part would have to report
Outcome classes leave four entries of the addition table open. Give each part its two stops as well and four inequalities settle 176 of the 369 open day-two sums, never wrongly — and every sum they leave has a bound sitting exactly on nought. Used as a lookup table instead, even the whole thermograph with the outcome beside it leaves 96 sums undecided. Four outcomes do what no thermograph can: who wins a day-two position alone, beside −1, beside 1 and beside a star names it, and no three questions do.
The pairs stand still and the order spreads
The share of pairs of game values the order can compare falls from 77.5 per cent on day two to 60 on day three and then stops: built twenty times under two recipes and calibrated, day four sits within two standard errors of day three. That looked like a floor under the order. Measured beyond pairs it is not. Chains of three keep falling, triples with no comparable pair keep rising, and the largest set of mutually confused values in a sample of 120 grows from about 20 to about 30 — every one of forty seeds wider than day three's mean. The floor belongs to one statistic, not to the order it summarises.
A universe grows by days, not by sums
A relation built from a sample of contexts can be repaired two ways: allow sums of more values, or allow values born a day later. On a sample of day four the first stalls — 147 pairs of different games stay merged from five summands to seven, through 25,696 contexts — while the 985 values of day three, a seventh as many, separate every pair. The reason is four lines long: the context that tells two games apart is always minus one of their options.
A foot is a menu, not a number
When a sum's fight ends level on nought, the winner is decided by what is left at the foot of each part's fight, and the obvious number to report from there is the atomic weight. Over 32,428 sums of a day-two and a day-three value it leaves 14,924 undecided, and on day two it does worse than a single outcome letter. What sits at a foot is a choice, made by one player, with someone to move when it is made. Report all three and every sum of two cold parts is decided; 3,164 sums are still not, and each holds a hot part whose own fight is not the one played in company.
Temperature
How much is at stake, measured. Thermographs, cooling, and why a player moves where the game is hottest.
What is at stake
Some positions both players are desperate to move in, and some neither player wants to touch. The difference is a number — how much the move is worth — and it turns out to be the most useful single quantity for deciding where to play.
Reading a thermograph
A thermograph is two walls rising from a number line, closing in as the tax on moving increases, and meeting where the position stops being worth fighting over. Everything about a position's hotness is in the shape.
Playing the hottest
Given several independent fights, play in the one with most at stake. The rule is simple, it is what strong Go players do without being told, it is provably close to optimal — and it is provably not optimal, which is the interesting part.
Cooling
Charge a tax on every move and a fight becomes a number. The height of tax at which that happens is the temperature — so cooling is not a technique for finding the temperature, it is what the temperature is.
The endgame, accounted for
Add up what each region is worth, then add the biggest thing at stake, subtract the next, and so on down. On a board of simple fights the result is exact — and the moment one region has a fight inside it, the account is out by a point.
Two hot fights that add to a cold number
The mean of a sum is the sum of the means, every time. The temperature is not — it is bounded by the hottest part and is often far below it. Two positions each worth fighting over can add to a plain number that neither player wants to touch.
The same fight, eight times over
The mean is not roughly what a position is worth. It is the number that eight copies of the position stay close to — and the theorem is that the closeness does not decay as the copies pile up. The gap stops at the temperature, and stays there for ever.
All ladders · Every object named here · The position index · Figures that play back · Search