Abandoning games as second player

[Sir Fluffy]

The currently top ranked arena player abandoned 11 of their last 50 games before the first turn, mostly when they're second player or occasionally as first player with a poor action card selection.

It seems like proposing to abandon a game is a viable arena strategy if you're second player.

Maybe something needs to be tweaked to reduce the first player advantage and to better balance action cards?

Or have BGA implement something to prevent you abandoning games too often?

[Sir Fluffy]

Someone let me know that quite a few players use this abandonment tactic and it's been going on for at least the last 2 years.

Some very rough calculations to see how much of an impact this has:

According to these stats the first player win rate for 350+ elo players is around 55%.

If you play 100 games, abandon 20 as second player. You play:

50 as first player, win 55% = 27.5 games
30 as second player, win 45% = 13.5 games
Total: 41 wins out of 80 games. So approximately an extra 40 arena elo (a loss changed to a win against an equivalent elo player) per 80 games played.

It's probably more with marine worlds since I think the first player advantage is even greater with the new university.

[ArtEntre]

It feels like you've got two topics merged into one thread...

1) Should BGA arena allow abandoning games based on initial game state? (Or a question you didn't ask but is closely related... what about abandoning based on opponent elo?)

2) Does Ark Nova have a significant first player advantage?

I imagine both have been discussed previously, but for the second I'm more aware of the prior discussions. The reference you quoted, pulling out specifically a 54.7% first player win rate (out of 128 games) is a rather small data set. The 95% confidence interval for that data is win rate 45.7% - 63.5%; so by commonly used definitions of statistically significant "proof", it can't be used to conclude that any first player advantage exists at all (you'd want 50% to fall outside the confidence interval.

A recent large tournament of top players actually had a "second player advantage" in its win rate. Again, it was a small sample size, so not statistically significant proof of a second player advantage, but it's an example of how small data sets can give highly variable results.

I believe a more recent look at a much larger selection of high-elo 2-player games found a first player win rate somewhere around 51%. I don't have the number of games for that statistic, so can't calculate the confidence interval on it.

Intuitively, I personally feel like first player should be a big advantage, but in terms of the statistics that have been made available nothing has conclusively shown that thus far (and arguably the look at the largest data set I've seen implies any advantage is quite small if it exists at all).

[ArtEntre]

I believe a more recent look at a much larger selection of high-elo 2-player games found a first player win rate somewhere around 51%. I don't have the number of games for that statistic, so can't calculate the confidence interval on it.

OK, going back and looking for more details on this, the data set was approximately 8,000 matches with average elo 619, which gave a first player win rate of 51.2%, which has a confidence interval something like 50.0% - 52.3%.

As a matter tangentially related of context, the winner's average elo in that data set was 640, with the loser's average elo being 597.

[Sir Fluffy]

I believe a more recent look at a much larger selection of high-elo 2-player games found a first player win rate somewhere around 51%. I don't have the number of games for that statistic, so can't calculate the confidence interval on it.

OK, going back and looking for more details on this, the data set was approximately 8,000 matches with average elo 619, which gave a first player win rate of 51.2%, which has a confidence interval something like 50.0% - 52.3%.

As a matter tangentially related of context, the winner's average elo in that data set was 640, with the loser's average elo being 597.

Thanks, that's very interesting! Do you have a source URL for that analysis? I'd like to read more about it

[Monkeyunit]

I know some high ranked players ask to abondon the game if the elo difference is huge.
The chance of the low elo player winning is slim, but the potential elo gain is small for the high rated player so abandoning is reasonable for both parties. First or second player has nothing to do with it in that case.

[yazavats]

Arena matchmaking is lacking for high rank players. Which I'm not, but I know what their grinding process is like. They don't want to play out the games where ELO/arena points gain is minimal. So are you sure you're not misattributing here? I.e., did you dismiss the games where the stronger player likely proposed to abandon due to ELO/ranking gap? Cause first player advantage is real, but not that big of a factor by itself. Either way it takes two to abandon, so it's up to the other player to judge whether the request is due to a bad position jitters or the aforementioned reason.

[ArtEntre]

Thanks, that's very interesting! Do you have a source URL for that analysis? I'd like to read more about it

The overall "analysis" occurred as discussion is the Discord server, so there's not a full documentation as easy to link to. The analysis I was looking at was done in this google sheets (specifically the tab named something like "600 vs 600").

Note that tab has two sets of data. I was referencing the "All" column with a 51% win rate. I believe the other set of columns is for 600+ vs 600+ elo games with an apparent 52%, but there are fewer games in that 600v600 data set, so the confidence interval is larger (dipping below 50% on the low end) and I consider it less conclusive regarding first player advantage even though it'd be closer to the ideal analysis in terms of even matchmaking.

[Sir Fluffy]

Arena matchmaking is lacking for high rank players. Which I'm not, but I know what their grinding process is like. They don't want to play out the games where ELO/arena points gain is minimal. So are you sure you're not misattributing here? I.e., did you dismiss the games where the stronger player likely proposed to abandon due to ELO/ranking gap? Cause first player advantage is real, but not that big of a factor by itself.

That's a reasonable thought. However, recently abandoned games were against players with ELOs of 250, 265, 422, 496, 562 so it's probably not based on ELO.

It looks like second player abandonment is a problem for 7 Wonders Duel too. See Collectively abandon game strategy abuse

Either way it takes two to abandon, so it's up to the other player to judge whether the request is due to a bad position jitters or the aforementioned reason.

I suspect the other players think it's a genuine request to abandon and want to be nice (which the majority of players probably are), without realizing their opponent is gaming the system.

[Sir Fluffy]

The overall "analysis" occurred as discussion is the Discord server, so there's not a full documentation as easy to link to. The analysis I was looking at was done in this google sheets (specifically the tab named something like "600 vs 600").

Thanks, that's interesting to read over! It looks like it doesn't have any Marine Worlds games. It would be nice to see if that makes a difference once there are enough games to analyze.