How long does it take for your games to be decisive?

In the interests of science, I carried out a study on how long it took for my games to be decisive. I took as my dataset all the even 19x19 games I played since the proper OGS winrate graphs arrived in December, which is 34 (pretty low .__.)

Then I took as my condition of decisiveness either 1. resignation or 2. one of the players achieves 99% winrate from OGS Leela.

These are the percentages of my games which were decisive within 100 moves:
Against DDKs: 100%
Against SDKs: exactly two-thirds (67%)
Against dans: 100%
Cumulatively: 79%

The mean average move at which my games were decisive was 76.

5 Likes

The dataset (it was tab-formatted in Notepad):

Game Dec. Reason Op. Rank

21762823 103 Me + 99% 7k
21762587 82 Me + 99% 11k
21736074 49 Me + 99% 10k
21723018 88 Op + Res 8d
21705637 38 Op + Res 3d
21704648 67 Me + 99% 5k
21692098 112 Op + 99% 8k
21667602 42 Me + Res 8k
21666488 26 Op + Res ?
21664745 42 Me + 99% 13k

21534871 61 Op + 99% 5d
21523986 87 Me + Res 3k
21523657 57 Op + Res 3k
21513332 43 Op + 99% 3d
21513210 90 Op + Res 4k
21493564 115 Me + 99% 6k
21493157 52 Me + 99% 5k
21492958 61 Op + Res 8k
21492542 62 Op + Res 1k
21491580 37 Op + Res 5k

21410952 102 Me + 99% 8k
21361460 161 Me + 99% 2k
21319688 67 Op + Res 2k
21319516 31 Op + Res 2k
21235253 82 Op + 99% 3k
21196062 87 Me + Res 11k
21115496 146 Op + 99% 6k
21098432 74 Me + 99% 10k
21086796 55 Me + Res 11k
21074452 170 Me + 99% 5k

21069923 65 Me + 99% 12k
21069156 50 Op + Res 2d
20975684 78 Me + 99% 6k
20975171 93 Op + Res 8k

3 Likes

Is this something that could displayed in gotstats? That would be great!

Well, the guy who runs gotstats is active in the OGS forum. I can’t remember his username though.

The name is @Chinitsu and she’s a lady, I believe.
(I didn’t say girl, did I?)

4 Likes

Duly corrected ^^

No I’m an (almost) middle-aged guy with a wife and a son. I’m very curious why so many ppl on OGS think I’m a woman though @_@

About the statistic above, average game length was one of my plan. But turn out it is only possible by querying each games individually, so for people with like 7000 games it would take hours. When I realized that I discarded the idea T_T

8 Likes

Well, your profile pic is a woman. And your name sounds pretty feminine to the Western ear, like Chidori (the artist of The Empty Triangle).

3 Likes

My profile pic is Katarina in League of Legends, and my handle is a big Mahjong winning hand (equal to Flush in poker but need 14 cards) :sweat_smile:

4 Likes

So sorry. I didn’t know any of them.

2 Likes

just limit your site to “last 1000 games” by default
people like me unable to use your site, it takes too long already
also statistics for accounts that started as 25k and (are dans now) makes no sense
There should be limit default option

I always assumed you were female because of the avatar. (Guess this is a good lesson on assuming things)

The username is awesome though. I am a huge mahjong fan, but I only recognize Chinese terms. Googled. 清一色 aahhhh! Another reason to give you all the friendship points. :heart:

1 Like

:open_mouth: just when I said “7000 games” some one with 7000 games in their account arrives
@_@

And it’s a pretty cool idea. Because games are queried last to first, I can actually display the statistics for recent games while waiting for earlier games to come. I’ll try that out.

Oh and for now, if you can be patient and wait until the end, I have an “Export” button on the sidebar where you can download the datas to your computer and later “Import” them, that way you can avoid querying old games again.

Let’s repeat this study a year later.

As a reminder, a “decisive move” is considered to be either

  1. an achievement of 99% WR from the OGS bot
  2. a resignation

A timeout is not considered decisive.

Only even 19x19 games are considered. For reference, we can assume my rank as 4k.

Date Op. Move Dec. Converted?
03-14 5k 47 me 99% Y
03-08 4k 57 me 99% Y
03-07 4k 82 me 99% N (timeout)
03-05 4k none (139 mv.) — timeout (me)
03-04 4k 97 me 99% N (timeout)
02-28 9k 68 op 99% N
02-15 4k 46 op 99% Y
02-09 4k 63 op 99% N
02-07 1k 49 op 99% N
02-07 4k 63 me 99% Y
02-03 2k 44 me 99% Y
02-03 4k 125 me 99% Y
02-02 6k 85 me 99% Y
02-02 8k 59 me 99% Y
02-02 4k none (88 mv.) — timeout (me)
02-02 4k 64 me 99% Y
02-01 4k 70 me 99% Y
02-01 1k 112 me 99% N
02-01 12k 33 me 99% Y
02-01 4k 87 me 99% Y
02-01 9k 27 me 99% Y
01-24 3k 86 me 99% Y
01-22 4d 80 op + res —
01-22 4d 69 op + res —
01-20 4k 216 me 99% Y
01-19 10k 57 me 99% Y
01-18 2k none (89) — timeout (op)
01-18 2k 70 me 99% timeout (op)
01-17 2k 89 op 99% Y
01-15 4k 87 op 99% Y
01-11 9k 23 me 99% Y

What can we conclude from this?

  • There were only two games in which resignation was made before either player achieved 99%, me vs 4d. This suggests that my decision was impacted by my opponent’s high rank.

  • There were only two games in which the decisive move was after (100), and they were both against 4k (ie. evenly-matched) opponents. This suggests that games I play against even opponents have the capability to last longer as even-ish struggles.

  • There were three games decided in less than 40 moves. These were against two 12ks and a 9k. This shows that I can sometimes achieve decisive positions very quickly against DDK / weak SDK opponents.

  • 18 of 28 decisive games were “decided” between (50) and (99). That is ~64.3%, just under 2 / 3.

Comparative studies, especially from DDK or dan players, would be interesting.

1 Like

I would posit that this is evidence that 99% is too low a number to be considered decisive :slight_smile: I’m sure 99.5% may yield quite different results, so what meaning can one really draw from an arbitrary threshold?

2 Likes

I had game today vs 3k where I was ahead for 84 moves in a row. I was ahead by 76.6 points max and had 99.9%.
and … I lost anyway. Not because simple errors. I just almost killed big group but it became bigger and bigger until it was out of control…

So if opponent has higher rank and resigned you never can be sure that you really won.

2 Likes

Do you count games twice (or thrice or more) if the lead changes after reaching 99%?

2 Likes

I didn’t for this study, but sometimes it does bounce from one 99% extreme to the other over the course of the game.

That’s much more true for 90%, though.