SSCAIT top 10 crosstables

Krasi0 passed Marian Devecka’s Killerbot to become #1 on SSCAIT on 18 August. A 100 by 100 crosstable is too big, but here’s a crosstable of the games played among the top 10 SSCAIT bots from 17 August to 27 September. The top 10 are chosen based on their Elo ratings at the end of the period. The top number in each box is the winning rate of the bot in that row against the bot in that column; the bottom number is the count of games. The overall column is the winning rate against the other 9 top bots; it doesn’t have to be closely related to the Elo rating, which is computed with all 100 bots.

since 17 Aug	overall	kras	Iron	Mari	Mart	tscm	tscm	Leta	Wuli	Simo	ICEL
krasi0	66.67%		20% 10	57% 14	100% 17	0% 14	55% 11	100% 7	100% 9	100% 8	100% 9
Iron bot	71.28%	80% 10		78% 9	36% 11	89% 9	64% 14	40% 15	100% 9	90% 10	100% 7
Marian Devecka	65.38%	43% 14	22% 9		100% 6	86% 7	58% 12	82% 11	80% 5	50% 6	100% 8
Martin Rooijackers	56.82%	0% 17	64% 11	0% 6		64% 14	78% 9	70% 10	100% 8	75% 4	100% 9
tscmooz	54.43%	100% 14	11% 9	14% 7	36% 14		50% 8	38% 8	43% 7	100% 6	100% 6
tscmoo	48.75%	45% 11	36% 14	42% 12	22% 9	50% 8		50% 8	0% 4	100% 6	100% 8
LetaBot CIG 2016	55.06%	0% 7	60% 15	18% 11	30% 10	62% 8	50% 8		80% 10	100% 14	67% 6
WuliBot	19.12%	0% 9	0% 9	20% 5	0% 8	57% 7	100% 4	20% 10		0% 8	25% 8
Simon Prins	31.94%	0% 8	10% 10	50% 6	25% 4	0% 6	0% 6	0% 14	100% 8		100% 10
ICELab	11.27%	0% 9	0% 7	0% 8	0% 9	0% 6	0% 8	33% 6	75% 8	0% 10

It’s amazing that Tscmoo zerg scored 100% versus Krasi0 during the period. It’s equally amazing that former champion IceBot has a 0% score against 7 of the other 9 top bots; apparently once you’re better, you’re a lot better. Bots have improved that much in the last 2 years.

Here’s the same crosstable, except starting 1 January 2016.

since 1 Jan	overall	kras	Iron	Mari	Mart	tscm	tscm	Leta	Wuli	Simo	ICEL
krasi0	45.47%		49% 57	26% 57	78% 64	13% 61	16% 73	100% 10	83% 35	55% 56	55% 51
Iron bot	53.37%	51% 57		41% 37	39% 66	58% 43	54% 59	40% 15	85% 27	55% 47	64% 50
Marian Devecka	63.12%	74% 57	59% 37		57% 100	72% 67	53% 95	79% 14	83% 12	80% 95	49% 106
Martin Rooijackers	53.71%	22% 64	61% 66	43% 100		57% 168	47% 230	64% 11	74% 31	44% 162	81% 164
tscmooz	62.23%	87% 61	42% 43	28% 67	43% 168		66% 142	27% 11	57% 23	77% 138	84% 128
tscmoo	62.43%	84% 73	46% 59	47% 95	53% 230	34% 142		43% 14	53% 17	88% 154	84% 161
LetaBot CIG 2016	56.64%	0% 10	60% 15	21% 14	36% 11	73% 11	57% 14		69% 13	100% 17	75% 8
WuliBot	24.63%	17% 35	15% 27	17% 12	26% 31	43% 23	47% 17	31% 13		16% 19	19% 26
Simon Prins	33.01%	45% 56	45% 47	20% 95	56% 162	23% 138	12% 154	0% 17	84% 19		38% 136
ICELab	33.73%	45% 51	36% 50	51% 106	19% 164	16% 128	16% 161	25% 8	81% 26	62% 136

The number of games in each cell depends on the lifetimes of both bots; not all were active over the whole period. The cells are mostly in paler colors, because many of the bots were updated during the year—they were neither always weak nor always strong.

I want to create a breakdown by map, but the data is not supporting it. There are 14 active maps. If I render data from the shorter period, there aren’t enough games for all the maps to be played in each pairing. If I use the longer period, there are usually enough games but the bots vary in strength with time so we’ll see a smear.

To break out crosstables by map, I’ll have to combine data one way or another. I could combine over time, using the longer period. I could combine bots by race, or try to compare groups of macro bots versus rush bots. I could lump together maps by number of starting spots. Or I could just do bot-map instead of bot-opponent-map. What do you think is useful? Maybe I should start with bot-map and then break it down further to bot-some group of opponents-map?

Trackbacks

No Trackbacks

Comments

krasi0 on Friday, October 7. 2016:

I believe I've earned most of my ELO via proxy (the two LetaBot versions that have drained it from most of the other top 10 bots).
Wrt the losses against tscmooz, I think it can be explained as follows: the BO that I used then was too vulnerable to one of the BOs (2 hatch mutalisk?) used by tscmooz and the latter has switched entirely to that BO against me (helped by the bot IO functionality). My plan is to come up with a counter to that in my BO arsenal at some point.

Add Comment

Name*

Homepage

Comment*

In reply to

E-Mail addresses will not be displayed and will only be used for E-Mail notifications.

To prevent automated Bots from commentspamming, please enter the string you see in the image below in the appropriate input box. Your comment will only be submitted if the strings match. Please ensure that your browser supports and accepts cookies, or your comment cannot be verified correctly.
CAPTCHA