Board Game Arena

FrankJones wrote: ↑12 February 2026, 19:25

FSKFSK wrote: ↑12 February 2026, 18:00 For team games, it sometimes counts as a tie with your teammate.

That could fit with the irregularities I observed.

That's asinine. Why would they calculate Elo that way for Spades? That's not an Elo flaw, that's just a terrible use of Elo.

From what I see from the Elo hover-over pop-up, it registers it as a tie with your teammate but then takes 100% off so that particular pairing is always a net of 0 on your Elo. What does this actually mean? No idea.

FrankJones wrote: ↑06 February 2026, 16:14
May I ask, why do you play players with much lower Elo than you, thereby putting your Elo at risk in a high-variance game?

So back to this. I play good players AND above meaning there are strong players mixed in there as well.

Whether my opponents are good or strong doesn’t matter. My point is that the probability calculations are off as a whole for luck games. Meaning it is inevitable for elite players to eventually fall to players inferior to them because every game is giving/taking incorrect rating based on false probabilities. Converging the entire field.

Frank, you gave an example of two solid players playing in a match and how 1 would receive 13pts for victory and the other 7pts (66/33) in a non luck game when you think that both could win. So imagine how damaging that same system is in a 4 luck game that is inherently closer to a 50\50 split.

An example was given about how these incorrect probabilities would affect players 400elo apart but why are we going to that extreme? If you take a 330 vs 300 elo player in 4 luck game they are practical even. But the 330 player is expected to win at 10% higher (55-45). At only 60 (360-300) elo difference it is 20% 60-40. That is extreme in a 4 luck game and if you can beat someone 20x more out of 100 in such a luck based game you are FAR superior.

BarnardsStar wrote: ↑12 February 2026, 20:05

FrankJones wrote: ↑12 February 2026, 19:25

FSKFSK wrote: ↑12 February 2026, 18:00 For team games, it sometimes counts as a tie with your teammate.

That could fit with the irregularities I observed.

That's asinine. Why would they calculate Elo that way for Spades? That's not an Elo flaw, that's just a terrible use of Elo.

From what I see from the Elo hover-over pop-up, it registers it as a tie with your teammate but then takes 100% off so that particular pairing is always a net of 0 on your Elo. What does this actually mean? No idea.

Previously you also lost/gained elo from your own teammate. Meaning before the game even started based on your teammates elo you were already up/down regardless of the outcome of the game. This was changed by bga last December I think (maybe it was 2yrs ) to at least tie with your teammate. Which is MUCH better system but it is still VERY flawed.

Frank being a better player than his teammate will receive less of the total elo won by his teammate in the game and also lose the majority if his team loses. It is very backwards. A clear example of what happens when you play in a system with inaccurate probabilities. & why so many spades players elo go all over the place.

ChiefPointThief wrote: ↑12 February 2026, 20:16

FrankJones wrote: ↑06 February 2026, 16:14
May I ask, why do you play players with much lower Elo than you, thereby putting your Elo at risk in a high-variance game?

So back to this. I play good players AND above meaning there are strong players mixed in there as well.

Whether my opponents are good or strong doesn’t matter. My point is that the probability calculations are off as a whole for luck games. Meaning it is inevitable for elite players to eventually fall to players inferior to them because every game is giving/taking incorrect rating based on false probabilities. Converging the entire field.

Frank, you gave an example of two solid players playing in a match and how 1 would receive 13pts for victory and the other 7pts (66/33) in a non luck game when you think that both could win. So imagine how damaging that same system is in a 4 luck game that is inherently closer to a 50\50 split.

An example was given about how these incorrect probabilities would affect players 400elo apart but why are we going to that extreme? If you take a 330 vs 300 elo player in 4 luck game they are practical even. But the 330 player is expected to win at 10% higher (55-45). At only 60 (360-300) elo difference it is 20% 60-40. That is extreme in a 4 luck game and if you can beat someone 20x more out of 100 in such a luck based game you are FAR superior.

We're both saying the same (or similar) things.

Ratings in luck-based games are going to be volatile. I don't consider that a failing of Elo. If anything, I might make the case that Elo adds additional information, specifically, it can help estimate how much luck versus skill is present in a game. Looking at the Elo ceiling for a specific game can give us a good idea of the luck involved.

If I see an Elo gap that under optimal conditions predicts a specific win rate, I take that to mean, "If we played 1000 games, I expect to win 1000*(win rate) times, but in a smaller sample of games, especially a single game, anything can happen."

Some may see this as a flaw. I don't. Obviously, opinions vary.

ChiefPointThief wrote: ↑12 February 2026, 18:18 No. Your comment makes no sense lol. I have a 18% better win percentage playing against tougher opponents over player b (& we both play 2p). Amongst SEVERAL other reasons that I listed why I am in fact a better player than player b. What are you actually arguing?????

How do you know that? Have you looked at their recent games and found out that they aren't doing as well as you against strong opponents? That seems unlikely if their elo is higher.

It seems to me that your 18% gap in winrate is due to games that they played a long time ago, when they weren't as good at the game, so all those early games bring down their overall winrate.

That metric just isn't very useful on its own, because it includes all the games you ever played on this website. It's like if you judged an athlete's level by counting the games they played in high school.

This is also why systems like elo need to work in a way that gives results less and less weight the further in the past they are. What happened a long time ago just isn't relevant when it comes to determining a player's current standing.

ChiefPointThief wrote: ↑12 February 2026, 18:18

No. Your comment makes no sense lol. I have a 18% better win percentage playing against tougher opponents over player b (& we both play 2p). Amongst SEVERAL other reasons that I listed why I am in fact a better player than player b. What are you actually arguing?????

Chiming in on your reply to tbhp.

If you have an 18% higher win rate against tougher competition than player b, you are most likely a better player than player b.

If you are currently rated at a lower Elo than player b, that's just short-term fluctuations and statistical noise. Player b is on a lucky streak and you are on an unlucky streak. The volatile nature of the high-luck games means this will happen sometimes.

You can
A) Get annoyed and frustrated about it and say the Elo system is broken, or
B) keep playing, and your higher skill will get you back on a winning streak at some point and get you back near the top of the rankings.

Take a look at my stats for Race for the galaxy. My win percentage is close to 50% but slightly below. Yet I have a "strong" rating of 305. I have been as high as 350. Someone might look at these stats and say, "How can this mediocre player with a 49% win rate have a 305 Elo?

Well, I set my filter to "strong and above", so all my games are against players rated 300 to 500. So, I lose slightly more often than I win, but my Elo reflects the fact that I am essentially holding my own against higher-rated players.

The same is true of Splendor. I fumbled along in the beginning, not realizing the sub-optimal plays I was making. Then I read some strategy posts and studied. Since then, my win rate has been very high, and my Elo reflects that. But merely looking at my overall win% does not reflect my current skill level, because it includes games early on in which I didn't know how to play optimally.

FrankJones wrote: ↑13 February 2026, 14:10 If you have an 18% higher win rate against tougher competition than player b, you are most likely a better player than player b.

They haven't shown that they do, they are just basing this on their overall winrate of 68%, while the player they were talking about must be at 50%.

Like I said, this doesn't mean anything. I myself have a 71% winrate in Harmonies, which is lower than some other players and is extremely misleading. It's due to the hundreds of real time games I played carelessly when I was new at the game.

Also, if you play a significant number of multiplayer games, this will also influence your winrate, since only first places count as wins.