K-Value Questions

[Zangoosed]

Not sure if this is the right place on the forum for this topic, but I guess I'll post it here anyway.

Does something seem wrong with the following:

I just came back to organized play this season after a 4 year absence.

My 2003-2004 Match Record - 37 Wins 17 Losses 4 Ties Modified Rating 1803.00

My 2008-2009 Match Record - 38 Wins 16 Losses Modified Rating 1660.23

I understand things have changed, but I'm just confused as to why the changes were made. I also don't understand how it can be such a drastic difference with an almost identical match record. So in short, I'm not trying to sound like a pain, but I'm really just someone whose been out of the game for a while and is looking for answers. Any help would be appreciated, thanks!

 

[ApachePrime]

The reason is because the tournaments you went to or the opponent's you faced were not as high ranked as they were previously. Both of these factors determine how many points you gain.

 

[Chairman Kaga]

My 2003-2004 Match Record - 37 Wins 17 Losses 4 Ties Modified Rating 1803.00

My 2008-2009 Match Record - 38 Wins 16 Losses Modified Rating 1660.23

I understand things have changed, but I'm just confused as to why the changes were made. I also don't understand how it can be such a drastic difference with an almost identical match record.

Because the points have little to do with your record. The point value you gain/lose from a match is a function of the difference between your rating and the opponent's. A win against an equally-rated opponent will award you half the K-value. A win against an opponent 100 points lower than you will award you virtually nothing. A win against an opponent much higher than you gives you closer to the full K-value. The reverse of these is also true -- losing to someone equally rated costs you 0.5*k, losing to a lower opponent costs you more, and losing to someone much higher than you causes almost no change in your rating.

So it seems, this year, you've probably played a lot of opponents much lower rated than yourself. You get very little gain on the wins and take a big hit on the losses.

 

[NoPoke]

2003-2004 all tournaments had the same K value (32iirc). In subsequent seasons K values were tapered. Last year doing well at nationals was crucial (K=44), less so this year. This season battle roads are of minimal significance to rating (K=4), so fewer of your 54 matches actually have much impact on rating if you win or lose.

There is a weighting effect too: Late season results matter much more than early season results.

37 wins out of 54 matches would suggest a +135 elo point margin over your opponents. ie a 1735 rating. What is your record at just CCs, States and Regionals?

 

[Zangoosed]

What is your record at just CCs, States and Regionals?

16-4 at BR's (3 BR's played/3 Top Cuts/1 win)

16-8 at Cities (4 Cities played/2 Top Cuts)

3-3 at States

Those are just approximations. I realize those dont add up to what my match record is. Seems as though I lost a good bit of points at States from what Im reading on My Match History. I have an even record but the people I lost to were all rated well below me. I guess that explains why my ranking is lower than I expected from my record. Just sux that i've top cutted in 5 of the 8 tournaments I've been to this season and my rating is still junk. Doesnt seem fair that a person can have 1 good showing at States and be garbage in the rest of the season and yet have a better rating than me.

This season battle roads are of minimal significance to rating (K=4), so fewer of your 54 matches actually have much impact on rating if you win or lose.

My point exactly. Whats the point of BR's at all? The prize support doesnt even justify the gas money to get to the tournament most of the time. And the K-value is terrible, so its not like youre playing for points to help your rating. The BR's I won gave me +10 points to my rating, 8 packs, and a victory medal. My point is that it probably cost me $40-50 in gas just to get there, so why should I ever play another BR?

And you get more prize support for judging the event than you do for winning the event. When was the last time you heard that the referees/umpires made more than the players in pro sports?

 

[bullados]

Doing very rough calculations, here's what I come up with...

Your BR record is +12 wins, which roughly equates to +24 points. EDIT

Your Cities record is +8 wins, which roughly equates to +64 points. EDIT

Your States record is +0 wins, which roughly equates to +0 points.

Very roughly (VERY roughly) I would guess that your rating would be somewhere in the high 1600s. Of course, that doesn't take into account the "quality" of your wins and losses. Were you losing a lot to people ranked below you? Especially at Cities and States, that can make a *huge* difference. Click on your rating in your "My Stats" page to see a detailed match-by-match breakdown of all the points you earned or lost during the season. That should give you the best information you can ask for.

(I'm an idiot :redface

 

[swell]

Doing very rough calculations, here's what I come up with...

Your BR record is +12 wins, which roughly equates to +48 points.
Your Cities record is +8 wins, which roughly equates to +128 points.
Your States record is +0 wins, which roughly equates to +0 points.

Oh no... If K-value is 4 as in BRs, you'll get approximately 2 points for winning opponent with almost same rating. When you face an opponent with same rating, winning possibilities are 50/50, and you'll get 50% of the K-value when you win.

So when Zangoosed has +12 wins in BRs, he gets about 24 points. In cites the record is +8, he gets about 56 points.

 

[Pajamas]

bullados, your numbers are twice what they should be. I'd expect a ranking of 1688, but since several of Zangoosed's wins were against me, it doesn't surprise me that it's a little lower.

 

[ApachePrime]

bullados, your numbers are twice what they should be. I'd expect a ranking of 1688, but since several of Zangoosed's wins were against me, it doesn't surprise me that it's a little lower.

His rating is impossible to figure out without his opponent's numbers. I had a great win at BRs, an opponent who was 300 points ahead of me, and only gained 2.2 points. You do not gain the max K-value when you win/lose. So Bullados's numbers are pretty largely inflated.

 

[Pajamas]

His rating is impossible to figure out without his opponent's numbers. I had a great win at BRs, an opponent who was 300 points ahead of me, and only gained 2.2 points. You do not gain the max K-value when you win/lose. So Bullados's numbers are pretty largely inflated.

I understand that. If you want to make a rough estimate, you should use half the K-value, which is what I did.

 

[Zangoosed]

You guys are missing the point. Im not refuting my rating or saying its incorrect. I'm trying to say that it seems like the system is flawed to me. Seems wrong that u get penalized points for losing in the top cut. It almost makes more sense to just make the top cut, and then drop.

And i'm also trying to find anyone who can give me a reason to play Battle Roads at all. The points arent worth it, the prize support isnt worth it, I guess you're just playing for "love of the game"?

And i counted my match victories in which my opponents rating was significantly higher than mine, 15 total wins. So roughly lil less than half my victories were above people rated higher than me.

So when Zangoosed has +12 wins in BRs, he gets about 24 points. In cites the record is +8, he gets about 56 points.

That about accurately describes my rating. I would be at about 1680 if I had not played States. At States even though I went 3-3 and then dropped, I ended up losing about 17 points due to the unlucky losses to opponents rated lower than me. What can I say? Machamp donks are Machamp donks.

 

[elekid_957]

BR's are pretty much testing for natz, that's it. all you get is a nice victory medal if you win. i went 7-0 on one day and gained 5 points (or less) then went 3-1 twice and lost points.. i agree that the system is flawed, where you need to go undefeated to gain points.

 

[Pajamas]

I don't think comparing absolute rankings from year to year is necessarily any indication of a flawed system without more context. You could divide all the K-values by four, and you'd still end up with a ranking where the most consistent players are at the top.

BRs are intended to be low-risk, low-reward events that hopefully provide a preview of the nats/worlds metagame. In the end, most everybody is just playing for love of the game anyway... only the Fulops and Holtons of the world can reasonably expect come out materially ahead.