The new ELO-based ranking system

[gamerman01]

Very well said, happy to have 2 respected players weigh in.
I’m going to wait until MrRoboto can answer some of these (I don’t know why ELO changes are the same for victory and defeated after game 1)

Part of the answer is that at this point the sensitivity ratings were somewhat arbitrarily put out there by MrRoboto, and I somewhat arbitrarily changed them to see what would happen.
That is, on the data sheet to the right, we currently have 110, 90, 70, 50 at threshholds 3 games, 6 games, 10 games (3 and 6 of course have become accepted league threshholds, 3 for getting a firm tier, 6 for qualifying for BM)

The matter of a new player getting a reasonable seat at the playoff table after finishing minimum # of games for his very first year is fairly easy to address, I think.

You (plural) raised other points and every sentence is appreciated. I do not consider this post here a complete answer, but some thoughts in response to some of your thoughts that I think will help advance the discussion.

[MrRoboto]

First of all, I can’t quite follow @Arthur-Bomber-Harris last post. Sorry, but I don’t understand what you’re trying to say there.
All I can say is, that starting ELO is the same for everyone.

Then the easy answer to @oysteilo first post:
Winner and loser points may differ, when one of them (or both) has less than 11 completed games.

This is because the first couple of games weigh more than later ones. I accomplish that with a “K Factor”, which expresses the amount a game is worth.
That K factor is really high for the first couple of games and then gradually decreases. As @gamerman01 said, the exact values are chosen “arbritrarily”, although of course we thought long and hard about them.

That K factor is essential, because new players all start at 1500, which is almost guaranteed not a perfect rating for them (most newbies are worse than that average, but some might be a lot better too). So we need the system to move new players as fast as possible to where they belong. Normally, a game between two equals awards only 25 (or -25) points. If a new player is actually a 1900 or a 1100 player, those 400 points climb or descend would take a long time without that sensitivity Factor.

Right now, the K factor we settled on is:

As you can see, the first 3 games are worth a little bit more than double the later games.

We can talk about these values, they are not set in stone. But I want to emphasize the importance of bringing new players to their appropriate rating asap. The only alternative would be placement matches - this would mean that new players are not rated at all for their first 5-10 games and opponents would receive / lose just a fraction of the normal worth. I don’t like this option for us.

---

Now the other question concerning playoff ranking / seeding.

@AndrewAAGamer is correct, originally I planned to use the current year only for playoff seeding. He said he also prefers this and compared it to football, basketball, baseball and hockey.

Now that comparison has one gigantic flaw however: All of these sports are in a league system where every participant has a fixed number of games and the exact same opponents. So it makes sense to start a season with a clean slate.

However, this is not the case with our community. With OOB and PtV, players need to have 3 completed games, with BM4 they need 6. But 3 games are not nearly enough to properly rate a player, especially not if one of those games was an upset (an unexpected loss / win). If we had an entry requirement of 10+ games per year, I’d definitely go for a clean slate every Jan 1!

Our game is more like the other two sports you mentioned (golf or tennis), even though they are still a bit different: It’s impossible to burst onto the scene as a complete nobody and expect to participate in the biggest tournament with only 3 games completed. A first time participant of a Grand-Slam-Tournament has proven himself/herself over many matches beforehand in smaller tournaments. We don’t have that luxury.
I think our sports can probably best be compared to boxing, where everyone chooses their opponents and some have only very few matches per year, while others have some more.

I can change the system to rank playoff seedings only according to results in the current year. Which would mean everyone starts at 1500 (for playoff ranking only). But do we want that?

@oysteilo actually gave a great example!

Oysteilo started the year with OOB rating of 1669
He is 2-1 this year, with both of his wins being almost worthless (against dawgoneit), giving him only +3 each. He lost once against the #1 AndrewAAGamer for -16. Which gives oysteilo a final rating of 1659. Which is more or less the same rating he had for the last 7 years.

ArthurBomberHarris started the year with OOB rating of 1542
He went 5-0 this year, although only one of his opponents was really strong (he defeated #1 AndrewAAGamer!). He gained 110 in the process, which means he significantly improved his rating from 1542 to 1652. This is the highest OOB rating he ever achieved.

They are now almost identical in rating (1652 and 1659), with 32 or 41 total completed games. Which means the rating is very reliable, those two players are very likely extremely similar in strength.

Now it’s a personal decision: Do you think oysteilo should get the higher seed because of his slightly higher rating and the fact he mainainted roughly that rating for 7 years? Even though his rating basically stagnated this year? Then the system should stay as it is.

Or do you think ArthurBomberHarris should get the higher seed because he is on an upward trajectory this year? Remember, Arthur is probably not better than oysteilo (they are most likely equally strong right now) and it is the first time he achieved the same level as oysteilo. But if you think the improvement he showed this year is worth more the system should change to let everyone start at 1500 on Jan1.

Do you think AndrewAAGamer who sits comfortably at #1, with only farmboy being SOMEWHAT close should get the top seeding? He went 7-2 and increased his OOB rating from 1798 to 1830.

Or do you think the 5-0 of Arthur and the 4-0 of Booper this year is more impressive and should give both of them a higher seeding, despite both of them definitely being not as good as Andrew?

My personal preference is the first option, which is currently implemented.
But I can absolutely understand if you value recent results higher than overall strength! This is a system that should be backed by the majority of the community so please: WEIGH IN!

[oysteilo]

Well explained @MrRoboto . I support your recommendation here

[Arthur Bomber Harris]

I also support your recommendation of multiyear ELO for playoff spots in the playoff. I am clearly not the strongest player going into the tournament: I needed a fair bit of luck to beat Andrew. I am an average player over my career and did not improve this year. No need to give me the #1 seed.

My main comment is to have everyone who signs up have a spot in the main tier of the playoff instead of having an upper and lower bracket. That won’t be a problem for OOB this year, but give players a chance to get the crown as a medium guy can sometimes pull enough upsets to become #1. Perhaps more of an issue for BM playoffs.

[MrRoboto]

Good point, @Arthur-Bomber-Harris , I also support a single tournament tree instead of having different brackets. But maybe I don’t see the merits of having multiple brackets, perhaps @gamerman01 or someone else can enlighten me.

[oysteilo]

@MrRoboto said in Proposal for a new, ELO-based, ranking system:

Good point, @Arthur-Bomber-Harris , I also support a single tournament tree instead of having different brackets. But maybe I don’t see the merits of having multiple brackets, perhaps @gamerman01 or someone else can enlighten me.

The main reason to have brackets the way they are is mainly a time issue. The champion should be “crowned” before a new play off season starts. If I am not mistanken the BM champoinship is currently ongoing… Clearly, we wont have multiple playoff season going on side by side.

[AndrewAAGamer]

Excellent explanation @MrRoboto. You have convinced me.

[AndrewAAGamer]

@oysteilo said in Proposal for a new, ELO-based, ranking system:

@MrRoboto said in Proposal for a new, ELO-based, ranking system:

Good point, @Arthur-Bomber-Harris , I also support a single tournament tree instead of having different brackets. But maybe I don’t see the merits of having multiple brackets, perhaps @gamerman01 or someone else can enlighten me.

The main reason to have brackets the way they are is mainly a time issue. The champion should be “crowned” before a new play off season starts. If I am not mistaken the BM championship is currently ongoing… Clearly, we wont have multiple playoff season going on side by side.

I too support one bigger playoff per game version however, @oysteilo has hit the nail on the head. With an 8 player playoff format that means 3 games have to be finished to crown the champion. If we went to a 16 player playoff that means 4 games have to be finished. 3 games allows 4 months per game while 4 games only allows 3 months per game. While 3 months sounds like a long time with vacations and busy schedules it is not actually a full 3 months. Plus, assuming a Turn every 2-3 days, it takes at least a week to complete one full Round. 3 months means about 12 Rounds, not nearly enough for a tough game. Time is the factor.

[Arthur Bomber Harris]

What if there was a full tournament every two years, and a special format tournament with varying format and rules in the odd years. Such as partners, or DoW must be called on turn 1, or all bid units must be bombers or whatever is voted popular that year?

[AndrewAAGamer]

@Arthur-Bomber-Harris said in Proposal for a new, ELO-based, ranking system:

What if there was a full tournament every two years, and a special format tournament with varying format and rules in the odd years.

I would not be in favor of a full tournament every two years. We want to keep the community active. Having a tournament every year provides more incentive for people to consistently play.

[farmboy]

I think we should stick to 8 as well given the time issues. Its been pretty common even with 8 for the final (and sometimes more than the final) to not get finished until the following year.

I would just add that while I was concerned too that longer term rankings might not capture what has happened in the year, I’m being won over. And regardless, I think its worth trying this out and seeing how it plays out. As Roboto has explained, there is room to fine tune it if we do run into the issues. And once we see how it plays if over the next year or two, we can still revisit.

[farmboy]

I saw ABH’s last comment after I posted. I would still stick to the tournament as we have it, but nothing stops us from also doing a larger tournament every couple of years alongside league play if there is interest in setting it up.

[Martin]

I am also strongly in favor of an annual tournament with only one bracket. This should not be an issue with OOB and PtV in the near to mid-term future due to the relatively low number of participants.

We can work a lot with byes, with top seeded players only joining in round 2 or 3 and the defending champion only in the semi-finals, for example.

At the same time, we could make the days-per-turn rule stricter in the first one or two rounds of the Playoffs and say two days per turn for round 1.

Of course, you can still ask for a break in case of business trips, sickness or whatever, but this should shorten the first Playoff round(s) by a few weeks.

Thank you all for your work by the way!!

[gamerman01]

Lotta great ideas, this is what I was hoping for!
I read all posts but MrRobotos (will read that next)

But I can tell you right now, even with Martin’s cool ideas on speeding it up, that I am firm on a bracket of no more than 8 for a few major reasons.

Even with byes to reduce total number of games, it only takes one game to stop everything. Going to 16, even with some byes (which means less than 16 players) means many more chances for a weak link of the chain to hold everything up.

You can institute timers like the same as regular league games, but I don’t want the #3 player being bounced by the #12 or whatever in round 1 (assuming 1 and 2 get byes) because he took 4 days for a move.

Going with more than 8 also cheapens the regular season.

The thought of a big 16 man tourney is intriguing and interesting, of course, but there’s not much theater in the #3 mopping up the floor with the #14. Or the #1 to the #16 if there are no byes.

Personally I’d rather see a 2 round, 4 man bracket for the championship for the above reasons.

2022 BM playoffs
#1 is fighting #3
2nd bracket
#1 is fighting #2

2021 BM playoffs
#2 beat #4 for the champ
2nd bracket
#2 beat #4

2020 BM playoffs
#3 beat #5
2nd bracket
#3 beat #4

2019 BM playoffs
#2 beat #4
2nd bracket
#2 beat #1

2018 BM playoffs
#2 beat #1
2nd bracket
#1 beat #2

2017 BM playoffs
#4 beat #2
2nd bracket
#5 beat #7

2016 BM playoffs
#8 beat #2
2nd bracket
#7 beat #4

2015 BM playoffs
#3 beat #4
2nd bracket
#1 beat #2
3rd bracket
#3 beat #4

That is a LOT of actual results that show that almost always, #1-4 win it. The additional brackets give players #9 through #16 more competitive games and a chance to win a bracket and go into all-time league history.

In fact, looking at the data, the #5-#8 seeds are just there to take punches, fall to the mat, and make it a little longer and more interesting.

With brackets of 4, you get winners in September instead of February the next year.

I’m not making a decree and I’m not meaning to argue, just putting my position out there for you to think about.

[gamerman01]

Aaaand, there is another reasons for shorter playoffs than longer.

On the one hand, it makes matchups of players who otherwise haven’t played each other and maybe never would.

On the other hand, it forces players to play someone they didn’t choose. A HUGE part of league play is being able to choose your opponents, which, by the way, makes it VERY different than major sports where they have a set schedule to play 1-2 games against a wide variety of opponents.

Just more food for thought.
I need to read Roboto’s post now about sectioning years and whatever else.

[Arthur Bomber Harris]

@gamerman01 you left off the OOB results. ;). I am a mediocre player but have had a few big upsets in my career as I bribe the dice lords with lots of presents and saved-up karma.

[gamerman01]

@Arthur-Bomber-Harris said in Proposal for a new, ELO-based, ranking system:

@gamerman01 you left off the OOB results.

I did.
Not intentional, just didn’t want to scroll down lol

[gamerman01]

OK, now I have read MrRoboto’s post, and wow, is that eloquent!

I wholeheartedly agree with just about everything he said.

My favorite option is to use ELO at the end of the year for players who met the minimum games requirement and sign up (want in).

Since only 3 games (rightly) required for OOB and PTV, obviously you’re not going to have very accurate ratings in some instances. This has been true under the past system as well (uses averages, and clean slate each year), but now we have a lot more data.

A new player to OOB or PtV will rapidly get a fairly accurate rating (even in 3 games) playing against others who are known quantities. If starting everyone at 1500 at Jan 1, not as accurate.

I really like MrRoboto’s Oysteilo example. He only played 3 games and those 3 games really don’t give much information. He’s better than our beloved Dawg, but couldn’t beat AndrewAAGamer in one game.
This means he’s like most players. Somewhere between the top and bottom.
But with his lifetime rating of 1669 lowering slightly from these 3 games to 1659, we have a much clearer picture of the mad skillz Oysteilo brings to the table than the little sample of 3 2023 games.

[AndrewAAGamer]

@gamerman01 said in Proposal for a new, ELO-based, ranking system:

but couldn’t beat AndrewAAGamer in one game.

He came dang close to it…

[gamerman01]

@AndrewAAGamer said in Proposal for a new, ELO-based, ranking system:

@gamerman01 said in Proposal for a new, ELO-based, ranking system:

but couldn’t beat AndrewAAGamer in one game.

He came dang close to it…

Thank you!!
And that proves MrRoboto’s point and mine.

A loss is a loss and a system based on wins and losses doesn’t care how close a game is.
The lifetime ELO for him does show it.
With only 3 games, and a near miss against AndrewAAGamer, a year that stands alone merely shows a sterile:

2 wins at Dawg
1 loss to Andrew

Put that on your tombstone