Arena scoring question

[dschingis27]

ELO is doing exactly what it should do here.

No, it doesn't. I agree with you that it is not as easy as ChiefPointThief puts it but there are certainly better ways to have a competition in games with luck than the current format.

The problem is, the current system is even reinforcing the luck factor instead of mitigating it. Dear BGA, just doubling the K factor in Arena compared to normal Elo. Why, oh why, oh why??

I think Arena was not designed by someone with a love for competitions or competitive play.

[Ceaseless]

The base win/loss change in a two player game is +20 and -20, where the players had similar Arena ELOs to start with. As the starting gap increases then the initially higher player gets fewer for a win and gives up more for a loss, say +10 or -30, and the initially lower player the reverse, say +30 or -10.

It provides a brake on runaway ELOs into the thousands, which leaves lower ranked players thinking “why bother chasing”. It also provides a brake on the downside that someone can’t go to zero ELO. Typically 95% of players will be between 1300 and 1700.

And this is why it is a awful format. If you are outperforming the field all season long it should be a breakaway victory. This is the equivalent of playing a basketball game and the amount of pts the losing team gets for a basket changes depending on how much they are losing by. There are mercy rules but children don’t even compete w/ these type of conditions.

I have a lot of stats to suggest this is a bad system based off my own personal experience last season. To your first point if someone enters arena late they are starting off w/ low points. Imo to help w/ this problem elo or previous arena rankings should be considered. I beat the reigning champ and a player that had more elo than me several times but because they were still in the lower leagues I didn’t get much pts. Then they went went on a streak and beat me several times and because the system is flawed it awarded them double pts. They were able to stop playing after that and just stayed at the top. This was the only player now I could get fair pts for defeating. So I had to do 4x more work to achieve the same results.

If you say limit games to elite only that also doesn’t work. I defeated a top ranked player 21 of 28 times and lost arena pts to this player because the system awarded them anywhere from 3.5-6x more pts for their victories.

I don't follow this basketball analogy. Rating is changed for winning/losing games. So I fail to see how your scoring analogy makes any sense at all.

Your Arena examples sound like you just faced someone who didn't have high rating in the Arena, so you didn't get many points, which seems reasonable. Their rating outside of the Arena shouldn't have any impact on their Arena points, otherwise people who have high rating elsewhere would struggle to move up the ladder despite high performance.

[Ceaseless]

Just be exceptional and don’t lose. Or reach the top. Dont play anymore and hope no one can catch you.

I think this one really resonates with me. I think at the core is that we (me at least, but likely many other players) want to play arena because it’s fun, because it might attract opponents that want to play competitively, and it might even get you games where otherwise the game is very dead.

So then the current arena format doesn’t seem to have any of these in mind, or do anything particularly well in any of the directions.

I mean it does get you games, particularly at the start of a new arena season - I tend to encourage people to play at this time also which kind of reinforces it

But in the long run outside of the first wave or two of games, it starts becoming unenjoyable, not just for people at the bottom but at the top too.

I’ve topped the arena in a few games, but unfortunately the current method of using Elo in the elite leagues I think is actually detrimental in encouraging players to play. When you’ve hit number 1, but it will take you many wins to recover the points loss of just one loss, it’s not worth even queueing up and risking losing to what could be a top 10, top 5 etc player joining late, or just generally having to play another strong player.

The current format encourages you to get to the top and then only do the minimum required to stay there. If someone passes you out, play only as much is necessary to pass them out and stop again.

The problem then is that this can all echo down the arena ladder, if number 2 wants to catch number 1, they have the same dilemma of 1 loss needing to be countered by several consecutive wins, so maybe they’ll stop playing, or only play on a good day - maybe they’ll just try again next season.

I think it’s better to assume players want to play the game, but it poses a dilemma where one needs to weigh up wanting to play vs wanting to win and becomes a silly optimisation.

At the lower end of the arena and I’m there in many games too, after a time into the arena season, you can be in a situation where the only players playing are far above your rating and each arena game is a guaranteed loss. So you’re stuck on your bronze 1 all season while the elite players are risking a lot to gain little in these matchups.

Now to be fair, it’s not all downsides, if the matchups happen and with the right attitude you can always learn something from games against stronger players, but I think this part requires a lot of dedication to strictly improving and probably isn’t fun in a competitive sense because the games aren’t necessarily competitive.

So it won’t necessarily drive average or intermediate players into the arena unless they just want to play games to gain experience more so than a realistic chance of winning (I don’t even mean in the games I play or am somewhat ok at - there’s much larger Elo gaps in other strategy games with top elo players breaking a 1000 or so elo).

Solutions?
Do I know the solution? No not really. Some player pools are so small it can be impossible to get live games, difficult to get turn based games and who knows about arena.

Splitting small pools too much for pairing won’t make sense.

Lowering the potential loss however for players in higher leagues and positions should in theory make it more reasonable for players that want to keep playing to keep on playing. They won’t have to balance off 1 loss against multiple wins, they can just focus on naturally trying to have a higher w/l ratio while playing a lot of games.

One might worry whether it makes players uncatchable, but equally forcing all players to stop playing and the arena to freeze up with no matches also makes players uncatchable.

Hopefully some more ideas come up in future discussions anyway. I really like the concept of arenas and playing with them, it’s just that no matter whether I’m top or bottom of a ladder, it becomes demotivating to play like halfway through the season or earlier.

This doesn't solve anything, that's just rating inflation. Players just end up playing until they hit the new rating peak where they can't afford to lose. That's how ratings work in general. As you get ahead of other players, you lose more points to a degree that it cancels out higher wins. If this never happened rating would just be about whoever played the most games after a basic level of competence, as players with better win/loss ratios would still get passed by people just playing a greater volume of games where they win the majority of them. There has to be a ceiling where your losses keep you from rising unless you win an exceptional amount, and the better you are the further you get under those conditions.

[Mathew5000]

The current format encourages you to get to the top and then only do the minimum required to stay there. If someone passes you out, play only as much is necessary to pass them out and stop again.
...
I think it’s better to assume players want to play the game, but it poses a dilemma where one needs to weigh up wanting to play vs wanting to win and becomes a silly optimisation.
...
Lowering the potential loss however for players in higher leagues and positions should in theory make it more reasonable for players that want to keep playing to keep on playing. They won’t have to balance off 1 loss against multiple wins, they can just focus on naturally trying to have a higher w/l ratio while playing a lot of games.

I'm just quoting a few paragraphs from Shinuito's insightful post, I encourage everyone to read the whole thing and think about it carefully.

I can think of a few ways of creating a formula for Arena standings that reward games won while also incentivizing continued play. For example, how about: gain 3 Arena Points (AP) for a win, and gain 1 AP for a loss. That's a simple system, but maybe it's too simplistic. At a minimum you'd want to adjust it so that a win against a top player gains more, while a win against a bottom-ranked player gains less.

Here's another formula to calculate each player's arena points: 100*W/G + (G - M), where W is the player's number of wins, G is the player's number of games played, and M is the median number of games played by participants in the current arena season. For example, if there are 51 players who have participated in the current arena season for the game in question, you put them in order of number of games played, and then M is the number of games played by player number 26 on your list.

So if I've played 10 games with a 7:3 won-lost ratio, but M is 16, then my arena points for ranking purposes is 70-6=64. And someone else who's played 25 games with a 16:9 won-lost ratio would have 73 arena points, calculated as 1600/25 + 16.

That's just an idea off the top of my head for how we could develop a formula. This specific formula would obviously need several tweaks: first, we want to tweak it so a win against a good player is worth more against a win against a bad player. Second, we want to tweak it so that a lot of players signing up and playing just a few games do not artificially deflate M. (Perhaps in the definition of M, just exclude players who have played fewer than 5 games; they would not be considered when computing the median. Or maybe it would be better to calculate M as the median of the number of games played by players with a certain number of wins. There are lots of possibilities.) Also you need to tweak it so that you can't rise to the top of the standings by playing thousands of games and losing them all. Perhaps you have to take into account the standard deviation of number of games played, so that the "G minus M" part of the formula rises more slowly as G gets more than one SD above M.

[ChiefPointThief]

I don’t understand how ppl like this system for luck based games like 99. Someone’s whose played less games than you can come along and take all the points you’ve worked for in a compete luck based game

How do you "work for" points in a complete luck based game lol. And why would anyone play a complete luck based game competitively (or expect any meaningful ranking in those)?

I mean it might depend on what exactly you mean by complete luck based game. If the game really is like 100% complete luck based like I throw a dice you throw a dice, whoever rolls higher wins, then I agree there’s no real playing competitively.

However there’s games you might describe as “complete” luck based games, like backgammon, where you can only really do what the dice allow you, so you can only really win “by luck” in a manner of speaking. In another fashion though, you can make decisions based on the dice rolls that increase your chances to win more on average, and so too with adopting strategies that might lose some games, but will win more on average.

So you might come out of a “complete” luck based game with 60:40 win loss rate, or more 65:35, 70:30 - I suppose it depends on what the game allows for, your opponents etc.

Still though, if games are not weighted equally then having a 70:30 winrate might still net you a big overall loss in rating - which is again the point above that arena might encourage better players to quit while they’re ahead in games of no chance or otherwise.

Yeah. Maybe I should've said heavily luck based games but you nailed the response. Also I don't want to keep posting shinuito's entire original comment because it is lengthy but I agree with that as well.

[buu441]

Here's another formula to calculate each player's arena points: 100*W/G + (G - M), where W is the player's number of wins, G is the player's number of games played, and M is the median number of games played by participants in the current arena season. For example, if there are 51 players who have participated in the current arena season for the game in question, you put them in order of number of games played, and then M is the number of games played by player number 26 on your list.

So if I've played 10 games with a 7:3 won-lost ratio, but M is 16, then my arena points for ranking purposes is 70-6=64. And someone else who's played 25 games with a 16:9 won-lost ratio would have 73 arena points, calculated as 1600/25 + 16.

This one in particular is so bad. It's basically rewarding whoever has the most amount of free time and has nothing to do with skill at the game. Like after a 100 games played, just playing a game gives you more points than winning. Like sure, the median will also increase over time, but the top player will just be the one who has played the most number of games in the season.

[shinuito]

Also I don't want to keep posting shinuito's entire original comment because it is lengthy but I agree with that as well.

It should be possible to press quote and then still edit the quotes.

At least I remember in other software being able to do silly things like

This is not a real quote - the text has been edited.

But I agree my post was a bit long

[detlefchef11]

So, the more I think about it, the more the current system makes sense to me. You should absolutely gain less/lose more if you play against someone who is way behind you in the standings.

Case in point: I made my way to the top 5 in wingspan this season on the strength of a win streak of games that I had started when I was silver and gold so many of them were bronze and silver players. Then, once I got to elite, I started getting more games against better players. But your win rate against another good player is nowhere near as good as against a weak one. So, even if you only get 8 points or so vs a weak player, your paths to victory are so much more. You can more likely overcome a bad starting hand if your opponent doesn't know how to take advantage of their better start. Things like that. If you just need to get to 80 instead of need to get to the 90s to win...

So, if anything, you should get even LESS points for beating a lower ranked player.

[ChiefPointThief]

So, the more I think about it, the more the current system makes sense to me. You should absolutely gain less/lose more if you play against someone who is way behind you in the standings.

Case in point: I made my way to the top 5 in wingspan this season on the strength of a win streak of games that I had started when I was silver and gold so many of them were bronze and silver players. Then, once I got to elite, I started getting more games against better players. But your win rate against another good player is nowhere near as good as against a weak one. So, even if you only get 8 points or so vs a weak player, your paths to victory are so much more. You can more likely overcome a bad starting hand if your opponent doesn't know how to take advantage of their better start. Things like that. If you just need to get to 80 instead of need to get to the 90s to win...

So, if anything, you should get even LESS points for beating a lower ranked player.

Recently I played the #2 player and received 16pts. We are both elite and top of the field. We also only started 60pts away from each other. If he won he would've received 24pts. 150% what I got. This isn't normal standards for any competitive competition. There are favorites in every competition but if the underdogs want to win the competition they have to win fair and square w/o a boost. To the people saying I am playing inferior competition (kids at the park) that isn't the case. Because of this weighted system I won head to head match ups against top 3 players but lost overall arena pts. I beat a top 10 player 3x more than they beat me and lost arena pts. I beat the number 2 player 5 matches more and only gained 14pts when the net for 1 victory should be 20pts.

[detlefchef11]

So, the more I think about it, the more the current system makes sense to me. You should absolutely gain less/lose more if you play against someone who is way behind you in the standings.

Case in point: I made my way to the top 5 in wingspan this season on the strength of a win streak of games that I had started when I was silver and gold so many of them were bronze and silver players. Then, once I got to elite, I started getting more games against better players. But your win rate against another good player is nowhere near as good as against a weak one. So, even if you only get 8 points or so vs a weak player, your paths to victory are so much more. You can more likely overcome a bad starting hand if your opponent doesn't know how to take advantage of their better start. Things like that. If you just need to get to 80 instead of need to get to the 90s to win...

So, if anything, you should get even LESS points for beating a lower ranked player.

Recently I played the #2 player and received 16pts. We are both elite and top of the field. We also only started 60pts away from each other. If he won he would've received 24pts. 150% what I got. This isn't normal standards for any competitive competition. There are favorites in every competition but if the underdogs want to win the competition they have to win fair and square w/o a boost. To the people saying I am playing inferior competition (kids at the park) that isn't the case. Because of this weighted system I won head to head match ups against top 3 players but lost overall arena pts. I beat a top 10 player 3x more than they beat me and lost arena pts. I beat the number 2 player 5 matches more and only gained 14pts when the net for 1 victory should be 20pts.

Now THAT doesn't make much sense. Unless you were way ahead of that player in Arena points, it should have been nearly even or even in terms of what you were both staking on the match. My point is, it seems like the spread is not enough in games where one player is ranked much higher than the other. I'm currently in the top 50. If I play against someone in bronze, with an ELO barely high enough to qualify for the arena, a bunch of things have to. break against me in order for me to lose. Well, I played someone who had obviously played the game a lot, just not here, and they played a great game. But, for the most part, my win rate against those players is about 90%. Yet, the delta in points gained vs playing someone that I'm equally ranked with pays me as if my win rate is more like 66%. So, with that in mind, it seems the best way to climb the ranks is to get fat against lower level players.