True Skill Rating (TSR) mod v.4.4 (01/20/12)

[[DARE]DarkZero901]

@ Zaraki thx 4 the quick info. yes i ve dL ed the mod update and i also guess it s lowered cause of the wards formula another question is This Team TSR winnig under the balance thing. is it bugged coz lasts games balance percantages r stayin same after i join another game... And when i move up my cursor al players in the new game i join it made just only a little difference... i ll put up a screenshot this post soon...

Sadly, the system used to refresh/retrieve stats for the player requires you to hover over the area where their stats are shown. So the win% won't properly update until you've hovered each new players stat area and retrieved the correct TSR value. I can look into trying to find a better way to update stats.

hey,

my trs stats are a bit bugged...in MM it looks often like this:

dunno whats up, but the other ones HAVE TSR (so they played pubbs and the tsr stand in their stats) but its not shown

I was pretty sure that the mod doesn't work with MM. If by some chance you do get stats in the MM, try updating them by hovering over the area where TSR should be and over the little circle that has the stats button. Might work, might not.

[[DARE]DarkZero901]

The thing is, while he can, he SHOULDN'T.
A Dsham can easily get 150 CK after 30 Minutes, but there are only so many Creeps on the map;
He will push all 3 Lanes, making his Carries unable to farm and essentialy losing the game for his Team.

There ARE exceptions (for example: Pollywog that wants a really fast SotM-Refresher),
but again: These are somewhat semicarrying then, and shouldnt buy Wards (or at least not many).

Very very basic Formula: 5 starting TSR
Currently:
- +2 for high CK
- +1 for good CK
- +2 for high Wards placed
- +1 for good Wards placed

Someone who plays mainly Carry will get +2 = 7 TSR.
Someone who plays mainly Support will get +2 = 7 TSR.
Someone who plays a mix of both will get +1 +1 = 7 TSR.
Someone who plays Support that farms too much will get +2 +2 = 9 TSR.

My proposition:
- +2 for high CK
- +1 for good CK
- +0.5 for high CK when Ward stat is also high*
- +2 for high Wards placed
- +1 for good Wards placed
- +0.5 for high Wards placed when CK is also high*

Someone who plays mainly Carry will get +2 = 7 TSR.
Someone who plays mainly Support will get +2 = 7 TSR.
Someone who plays a mix of both will get +1 +1 = 7 TSR.
Someone who plays Support that farms too much will get +2 +0.5 = 7.5 TSR.

*Of course not taking both at the same time, but the higher one stays as +2, the lower reduced to +0.5

That was my basic idea. But wait with it for now, lemme check on the current numbers ;b

I like the suggestion but I'm trying my best to stay away from if statements with the formula. How long that will last, I don't really know. For now, I put much less weight on wards because the server doesn't handle wards properly and abusing the ward statistic is way too easy.

__________________________________________________ _________

Also I am still not quite sure on why your numbers are so strangely spread,

So with 5 being the <deadaverage> player, we go -1.1 for the "low average" and +0.9 for the "high average"?
And average <of the whole player base>? That'd mean someone with 1700 is a "good player".. which is imo not the case ;b
I'd prefer a farther spread (like 0-15 instead of 0-10) anyways, as I do not think the current one can really show us 'minor' differences*

*Ex: between high lvl Pubstars & competitive players

I know I say the average player is 5 but in reality, the formula comes out to have the average player at 4.9. So 1 TSR either way will create the bell curve we all love. Also, TSR won't show the difference between pubstars and competitive players (Chu). Competitive players won't have great statistics. Their KDR/ADR, denies, creep kills, etc will all be within reasonable ranges. Why? Because they play against opponents with the same caliber of skill (as mentioned below) and they don't play to get kills. They play to win. But competitive play is a monster of its own and we don't need to worry about that.

__________________________________________________ _________

The main problem this STILL has (and, in it's current form, will always have),
is that it doesn't take into consideration your opponents strength.
A high level competitive player that only plays scrims, tournaments, and 1800+ Inhouses will come out as an
"Average player" (4-6 TSR) with your rating system. Obviously: Dead wrong.
Same the other way round; a decent player that spends his time stomping noobs will come out as an
"Superior player" (8+ TSR), which again is dead wrong; both of these cases leading to harsh confusions.

Imo it REALLY needs to take PSR into consideration.
Average Stats but 1800+ PSR should at least make you a 7,
High Stats but 1500+ PSR should at best make you a 5.
Of course, PSR got it's flaws, but most of these can be corrected by looking at the Stats at the same time;
which your TSR essentialy does. Someone with 50 Games played, 80% Win and 1550 PSR should not type out as 9.0 TSR.

You are right. TSR does not and will not take the opponent's strength into account. It's not a rating system that's supposed to predict who will win a fight, it's a rating system of the individual player. However, I did implement a win% based on TSR to see how accurate it really is. This takes your opponent's strength into consideration.

Honestly, there is no point in integrating PSR into TSR because 1) PSR depends a lot on luck of heroes/teammates and 2) PSR is readily available to be used in conjunction with TSR.

You might argue that TSR also relies on luck of heroes/teammates but i disagree. I've been in many games where we are getting our asses handed to us but I maintain decent denies/KDR/ADR. My exp/min, creep kills, and gold/min are lacking during those games but by playing smart you can prevent getting a 1-10-2 score and ending up with a 1-2-2 score, which really isn't too bad. This shouldn't affect your TSR too much, but if PSR was taken into consideration, it would affect your TSR even more.

[[DARE]DarkZero901]

One more thing about including PSR in TSR: I've seen 1700 players with absolute **** stats (thus low average TSRs) and ended up realizing that they have no business being 1700s.

Sometimes noobs can have a 1700 PSR and still be horrid in game. While pros can be having a tough time with noob teammates and end up in 1500/1400s PSR. I want to be able to weed these players out and spot the noobs among 1700s and spot the pros among 1400s.

[[xKSK]w4rLord]

Screenshot about Team TSR win percentage bug

[netygraph]

Can someone post a screenshot of this in use?

edit: Should have checked the last post...

[[DARE]DarkZero901]

Screenshot about Team TSR win percentage bug

Damn I see what you mean. I'm gonna try replicating this and see what I can do about it. Thanks for reporting it.

By the way, it should have been Legion: 49.2% and Hellbourne: 50.8%

[[DARE]Zaraki]

It's great that some people are responding to this and helping with the tweaking of this mod, thanks a bunch; I know DarkZero901 also appreciates it. Some quick remarks.

First, TSR is supposed to help gauge a player's skill. As such, I don't think it will ever be very accurate in predicting what team will win, since a lot of times that has to do with other factors as well, such as hero picks and communication skills (if a team is composed of players who speak different languages, it is that much more difficult to coordinate ganks or coordinate counter-pushes, etc.)

Second, at this stage, TSR is not supposed to replace PSR but is supposed to compliment it. As I previously stated, you can have a low PSR but be a very good player, and vise versa. Example: You are a good player, but you have a string of 6-7 games in which you lose because one of your teammates leaves, is a noob, doesn't help the team, etc. That can drop your PSR by 100 points or more. Likewise, you are a bad player, but have a string of 6-7 games in which you win because the rest of the team is very skilled, you get good heroes picks or the rest of your team plays well together, etc. That can raise your PSR by 100 points. In either situation, your PSR is not reflective of your actual skill.

Finally, this is a work in progress so it is still not as accurate as DarkZero901 wants it to be, but with your inputs and suggestions, it is getting there.

@ netygraph You should download it and try it If you don't like it, you can delete it, just make sure your tell us why you don't like it

[[FEED]Solav]

chance of keeping the latest used formula in the OP?

[Yakuza]

Hmm sometimes I don't get the formula.
A guy with 50% winchance, but 1.3 k/d and 100 average creepkills was rated way lower then me.
It kindof works sometimes, but sometimes it's really weird.

[[DARE]Zaraki]

If you see oddities, please provide a screen shot or at least give some detail as to the name and the displayed TSR of the person. If providing names is against forum rules, please let us know so we can try and figure something else out. However, as the name isn't given to put someone down but rather to determine whether the formula is doing what it is supposed to, I don't see a problem.

[[N]Caeous]

Excellent program.

However, I would advise against putting any kind of negativity with regards to the leave percentage. A player might have decent skill but leave games. If I end up playing with him, then I want to know his skill level, not whether he leaves or not. I can check that myself in his stats.

If I'm playing in a No Leavers game, then leavers are out of my mind. I understand that this means that the TSR difference will also be lower, but the fact that it changes at all is deceiving. The simple fact that they might have Internet issues is a good enough reason to ignore this stat completely.

Also I just have to enable Custom 4 in your program to see that.

TL;DR imo whether you have left some games or not does not change your skill level.

[[DARE]Zaraki]

Completely agree with you. I think we almost have DarkZero901 convinced to take it out of the formula. Hopefully he will soon.

[DragooNick]

Excellent program.

However, I would advise against putting any kind of negativity with regards to the leave percentage. A player might have decent skill but leave games. If I end up playing with him, then I want to know his skill level, not whether he leaves or not. I can check that myself in his stats.

If I'm playing in a No Leavers game, then leavers are out of my mind. I understand that this means that the TSR difference will also be lower, but the fact that it changes at all is deceiving. The simple fact that they might have Internet issues is a good enough reason to ignore this stat completely.

Also I just have to enable Custom 4 in your program to see that.

TL;DR imo whether you have left some games or not does not change your skill level.

I like that Leaves/Kicks are in the TSR Formula, since i played now not to rarely with ppl who got 2,7KDR / 2ADR and not too high leaver Percentage [i didnt have this mod at that time] as those ppl were winning they stayed.
But as 1 Time my Brother and me decieded to **** his stats
He was gone faster than every noobleaver i have ever seen ...

Thats why i think its ok to let it influence the TSR, because if you think
hes a really good player because of his stats and he gets owned,
your so fast 1 ppl less in team .... :S

The influence should not be the biggest but prevent that ppl
with pubhunter stats + leaves are considered above average


@DarkZero901
I also got a bug i cant prove with a screenshot ^^"
everytime i color my "TSR|" in the options like that "^222TSR|" it works but after i restart the game its back to the standard "TSR|" in white :S

[[ASQ]voidSkipper]

I'd like to make a suggestion.

I haven't really looked at the capabilities of the scripting language used in these mods, so I'm not sure if this is possible but.

Would it be possible to modify it so it only counts the last 20-40 games?

The rating is great, but becomes less useful as time goes on because the longer you play, the less your improvement is reflected in your stats. After just 50 games, it becomes extremely difficult to move your average values - especially with stats like wards/game. If you only learn to buy wards after 90 games, but you've been buying 10 a game for the last 10 games, this should be reflected in your rating - not the fact that you've only bought an average of 1 a game.

[DragooNick]

I'd like to make a suggestion.

I haven't really looked at the capabilities of the scripting language used in these mods, so I'm not sure if this is possible but.

Would it be possible to modify it so it only counts the last 20-40 games?

The rating is great, but becomes less useful as time goes on because the longer you play, the less your improvement is reflected in your stats. After just 50 games, it becomes extremely difficult to move your average values - especially with stats like wards/game. If you only learn to buy wards after 90 games, but you've been buying 10 a game for the last 10 games, this should be reflected in your rating - not the fact that you've only bought an average of 1 a game.

I dont think this is possible and I doont even like the idea.
It will improve your stats, maybe not fast but it will,
if you learned placing wards after 90 Games then ofc you should
have to play another 90 Games to get that stats a bit better.

Also showing just the last games would allow pushing unreal stats ...


Edit: I just found a bug which i can prove


That calculation isnt rigth?
And is it me or arent all of the options displayed? The bar in which the informations are displayed stays that big and if I enable more options they arent showed, the are cut down
Just like the "T" from "TSR" from DaHu4b.

[Lust]

Another example of broken TSR comparison. This time it's even a 4v5. I hope you get this to work as intended as I like this mod .

[[DARE]DarkZero901]

Can someone post a screenshot of this in use?

edit: Should have checked the last post...

I will post a SS in the OP soon.

chance of keeping the latest used formula in the OP?

I will also post that in the OP soon.

Regarding the leave percentages; I see both sides of the argument and was almost ready to get rid of the leave percentage from the formula. However, since DragooNick said something about keeping it, I'm gonna hold my stance for a little longer .

I understand the concern with all the stat/win% bugs and am trying to get to them. Thanks for reporting them guys.

[[DARE]Zaraki]

I'd like to make a suggestion.

I haven't really looked at the capabilities of the scripting language used in these mods, so I'm not sure if this is possible but.

Would it be possible to modify it so it only counts the last 20-40 games?

The rating is great, but becomes less useful as time goes on because the longer you play, the less your improvement is reflected in your stats. After just 50 games, it becomes extremely difficult to move your average values - especially with stats like wards/game. If you only learn to buy wards after 90 games, but you've been buying 10 a game for the last 10 games, this should be reflected in your rating - not the fact that you've only bought an average of 1 a game.

If you check a player's stats, they don't show and improvement until he has played enough games to raise his stats. This takes time but if the player has truely gotten better, the stats will evenetually show this. Likewise, once this improvement starts showing up in the stats, TSR will start to reflect it as well.

Besides, I seriously doubt that you can pull stats from each game separately, run a calculation to average the stats out up to a certain amount of games, and plug that information into the formula. Even if it is, I would think it would be difficult to mod.

[[ASQ]voidSkipper]

I dont think this is possible and I doont even like the idea.
It will improve your stats, maybe not fast but it will,
if you learned placing wards after 90 Games then ofc you should
have to play another 90 Games to get that stats a bit better.

Also showing just the last games would allow pushing unreal stats ...

Why? It's supposed to show a true measure of skill, not a measure of whether or not the player has paid their dues after improving.

[VirtuosoZ]

Excellent program.

However, I would advise against putting any kind of negativity with regards to the leave percentage. A player might have decent skill but leave games. If I end up playing with him, then I want to know his skill level, not whether he leaves or not.

TL;DR imo whether you have left some games or not does not change your skill level.

Completely agree with you. I think we almost have DarkZero901 convinced to take it out of the formula. Hopefully he will soon.

This.