OT: Player rankings [PROTOTYPE IS UP!][Goalies added]

[elkranio]

Important update:

Prototype is up. You can now grade stats.
You need to be logged in to do this. Registration is very simple, no email confirmation needed.

In order to grade a stat select a player and then click on a stat you want to grade.

Link: http://icegrades.com/
Your feedback is very appreciated.

---- end of update ----

Hey guys, I have something on my mind that bugs me and I want to spit it out at last.

I was always wondering why there are no resources that have NHL player database with their skills rated. You know like they are rated in EA Sports NHL games.

Something like this: Skating: A+, Slapshot power: C-, Stamina: B and so on.

I am reading books on web-development at the moment (something like a hobby to me) and I was thinking about creating a site that will allow people to rank players and their skills. Of course ratings are subjective, but this site will calculate average values, so this should even things s bit.

I think it might become a very useful tool. For example you want to find a replacement for Oshie. You can select every RW in the league with their overall skill level above certain point and a "playoff performer" stat higher than average

I will probably be able to code all this stuff myself (though I am still learning), but I really want to know what do you guys think about it. Maybe this idea is stupid or just plain ****** and I don't see it.

And if this is a good idea I will definitely need some help (stat list, database management etc).

So what do you think?

-------------------
UPDATE #1:
Okay, so I've made some kind of a draft page just to see what it can look like.
Screenshot
I need your advice regarding this skill list. Maybe you want something to add or remove from list? Right now it is basically a EA NHL rip-off

UPDATE #2:
Dev server is available. You can check my progress here: http://isnifer.questworld.ru:8888/ (don't forget eights). Expect bugs.
Not much there at the moment, I am doing prep work mostly.

Click to read more...

 

[KirkOut]

it sounds like a good idea. my first concern is that if you leave it up to a vote/average, there will have to be some way to take into account that players from MTL/TOR/NYR/etc will be inflated due to large fanbases overrating their guys. But it's a cool concept overall.

 

[elkranio]

there will have to be some way to take into account that players from MTL/TOR/NYR/etc will be inflated due to large fanbases overrating their guys.

I was thinking about making a prototype first. Blues players only.
Just to see if the idea works out.

 

[KirkOut]

I was thinking about making a prototype first. Blues players only.
Just to see if the idea works out.

ok so just make a spreadsheet of 0's. Mission accomplished.

Which language are you using?

 

[elkranio]

ok so just make a spreadsheet of 0's. Mission accomplished.

Which language are you using?

A spreadsheet of Oshie-O's?
HTML/CSS/jQuery for front end, tornado and python for back end.

I haven't started yet, want to hear what you guys think about this.

 

[Dbrownss]

Cool idea.

 

[STL fan in MN]

I actually think that sounds like a pretty good idea. You'll probably have make some sort of adjustments to the scores (like throwing out the high/low scores), etc. There's also the potential of completely uneducated people (say, someone that truly considers Oshie to be an A+ skater) and people purposely being malicious to mess up your rankings (someone purposely rating something way higher or lower than reality just to screw up the average). Those are probably things you will eventually need to account for somehow but overall, I think it's a good idea.

 

[elkranio]

What do you think about the ranking system itself?
I am leaning towards letter based one.
Something like this: S - elite, A - top, B - above average C - average D - below average E - borderline NHL level, F - below NHL level.
Digital system may look good, but I think that letter-based one is much much easier to work with.

 

[elkranio]

There's also the potential of completely uneducated people (say, someone that truly considers Oshie to be an A+ skater) and people purposely being malicious to mess up your rankings (someone purposely rating something way higher or lower than reality just to screw up the average). Those are probably things you will eventually need to account for somehow but overall, I think it's a good idea.

Yeah, that concerns me a lot. But I hope it could be dealt with one way or another.

 

[2 Minute Minor]

Biggest issue I see is historical context. If you plug in a score for Gretzky, even though he's retired and not a good NHL player today, his numbers could look better than a current star player. You'd like to end up with something that is more useful for comparing contemporary players to each other in a given season.

 

[elkranio]

Biggest issue I see is historical context. If you plug in a score for Gretzky, even though he's retired and not a good NHL player today, his numbers could look better than a current star player. You'd like to end up with something that is more useful for comparing contemporary players to each other in a given season.

Yeah thought about that. Probably should save overall ratings once a season ends.

 

[EastonBlues22]

Biggest issue I see is historical context. If you plug in a score for Gretzky, even though he's retired and not a good NHL player today, his numbers could look better than a current star player. You'd like to end up with something that is more useful for comparing contemporary players to each other in a given season.

A weighted averaging system (votes accrued more recently, or in more recent seasons, are weighted more heavily) could help work around this. Suppose you could also cook up an aging algorithm that automatically lowers a player's historical votes (votes cast during previous seasons) as his age increases...new votes would either drag those results down even further if the player really sucks, or push him back up a bit if he's aging well, to naturally "fine tune" the process.

it sounds like a good idea. my first concern is that if you leave it up to a vote/average, there will have to be some way to take into account that players from MTL/TOR/NYR/etc will be inflated due to large fanbases overrating their guys. But it's a cool concept overall.

Can always exclude (or reduce the value of) votes cast from a player's current market (or city) from the tally.

Would have to be able to identify a voter's fanbase, either through mandatory self-reporting before they can vote, or maybe through IP location. If it's self-reported, you shouldn't tell them that this affects voting results, or they might intentionally mislead you. Probably should only let each exclusive IP address vote once a year, or some small limited number, if that can be managed.

Could also eliminate (or reduce the value of) votes from people who self-identify as being below a certain age as well, as there's probably more bias and less actual knowledge there.

 

[LUF]

Yeah, that concerns me a lot. But I hope it could be dealt with one way or another.

Have an internal system that weights the users. If someone is typically far from the scores other users give, then weight his ratings less. Also give more weight to the users who rated more players so to help prevent anonymous trolling.

 

[KirkOut]

A spreadsheet of Oshie-O's?

get with the times, man. surely you mean Boedker-O's

Can always exclude (or reduce the value of) votes cast from a player's current market (or city) from the tally.

Would have to be able to identify a voter's fanbase, either through mandatory self-reporting before they can vote, or maybe through IP location. If it's self-reported, you shouldn't tell them that this affects voting results, or they might intentionally mislead you. Probably should only let each exclusive IP address vote once a year, or some small limited number, if that can be managed.

Could also eliminate (or reduce the value of) votes from people who self-identify as being below a certain age as well, as there's probably more bias and less actual knowledge there.

Yeah, there could be lots of behind the scenes filters and weights to help make it more accurate. Another idea would be to only be able to rate one player at a time, and the player who shows up on your screen for you rate is chosen at random. This would prevent mass team or player-specific brigading. I'm picturing all the Ottawa fans rating Subban a 0 across the board while all the MTL fans rate Karlsson a 0 across the board at the first chance they get. This would also make sure that all players across all teams got more or less the same number of ratings.

 

[ManyIdeas]

hit me up if you need python help.

 

[STL fan in MN]

get with the times, man. surely you mean Boedker-O's



Yeah, there could be lots of behind the scenes filters and weights to help make it more accurate. Another idea would be to only be able to rate one player at a time, and the player who shows up on your screen for you rate is chosen at random. This would prevent mass team or player-specific brigading. I'm picturing all the Ottawa fans rating Subban a 0 across the board while all the MTL fans rate Karlsson a 0 across the board at the first chance they get. This would also make sure that all players across all teams got more or less the same number of ratings.

The main problem I see with the player you rate being chosen at random is that that would likely result in quite a few uneducated rankings. I would only want people ranking players they were familiar with. If the 4th line guy whose name I barely even recognize of some EC team I don't follow all of a sudden shows up on the users screen, that's not going to be a useful rating.

I suppose that could be mitigated somewhat by there being an option to skip that player (move on to the next random player) but I think it would still result in quite a few uneducated ratings.

If there are some players that have very few ratings then I would imagine that an ad could be developed to show right on the site targeted to users of certain characteristics (they've rated quite a few players from that player's team, their ratings show in the algorithm to be pretty accurate) to try to prompt them to rate the player that needs more data points.

 

[KirkOut]

The main problem I see with the player you rate being chosen at random is that that would likely result in quite a few uneducated rankings. I would only want people ranking players they were familiar with. If the 4th line guy whose name I barely even recognize of some EC team I don't follow all of a sudden shows up on the users screen, that's not going to be a useful rating.

I suppose that could be mitigated somewhat by there being an option to skip that player (move on to the next random player) but I think it would still result in quite a few uneducated ratings.

If there are some players that have very few ratings then I would imagine that an ad could be developed to show right on the site targeted to users of certain characteristics (they've rated quite a few players from that player's team, their ratings show in the algorithm to be pretty accurate) to try to prompt them to rate the player that needs more data points.

in this system I would implement not only full player skips but also the option to skip individual attributes. Say I'm rating Alexander Semin. I want to give him a top score on wrist shot, but I have no idea what his agility is like. I would need to leave that one blank. Hopefully enough people vote so that all the blanks can be filled in.

 

[KirkOut]

also, the random part could only be the first phase. once enough preliminary data is collected, you can try to control people intentionally trying to mess it up by not counting votes that fall beyond X number of standard deviations, or whatever.

 

[elkranio]

Well that's a lot of ideas. All right I'll try to make some kind of prototype in a few days (I hope I'll have time).
But how do you guys feel about this letter-based system?

 

[ManyIdeas]

Well that's a lot of ideas. All right I'll try to make some kind of prototype in a few days (I hope I'll have time).
But how do you guys feel about this letter-based system?

Sounds like a winner if you do A+ A and A-, not just a single grade. more accuracy if done correctly.

 

[Lord Helix]

Love the idea.

You seem like a guy with the know-how to do something great, elk. Just don't steal my CapNerd app idea that ties in trades, advanced stats, fantasy scores, etc.

 

[KirkOut]

hey i have an idea, what if you expanded the concept to include stuff like trades, or even advanced stats and fantasy stats?

 

[WiscoBlues]

To adjust for people messing with averages, you could make some sort of feedback system where people can rate the accuracy of an account's ratings. If you want to see more objective ratings, you can filter to only include ratings made by accounts with over 90% positive feedback (or whatever arbitrary number you want to use).

 

[Meatwagon]

To adjust for people messing with averages, you could make some sort of feedback system where people can rate the accuracy of an account's ratings. If you want to see more objective ratings, you can filter to only include ratings made by accounts with over 90% positive feedback (or whatever arbitrary number you want to use).

This is a good idea, allow people to vote on the given values and if it gets a predetermined %, then you could add a "verified" type of title/banner.

Also as far as the statistical breakdown, it would be easiest to use a basic statistical avg, you remove the outliers above or below 3 deviations then calculate the mean and the mean would become that players letter grade. I do like the idea of going with an A+, A, A- grading skill. Or you could build it like GPA for schools. 3.75+ =A+, 3.74-3.25=A, 3.24-3.01=A-. Something like that?

Also maybe we could talk out some sort of ranking system that takes into account a players salary vs their production. So if a player is in the top 5% at their position for stats, but make the median avg for a typical top 6 player at their position, they get some sort of added value. Like "Player X is the best player for the money". This way we can take these rankings to the trade board for the Malkin to the Blues thread and once and for all prove that Malkin for Backes, Shattenkirk, Schwartz, Fabbri, Barbashev and a 1st is actually ridiculous, thus this poster gets shamed to the point that they no longer post!! I'm only mostly kidding Easton.

 

[The Grouch]

Yeah, that concerns me a lot. But I hope it could be dealt with one way or another.

IMDB has a weighted system for their voting feature. They "don't disclose the exact methods used, to avoid leaving the scheme open to abuse."(their words), but they have a detailed explanation of the overall system that may be helpful...

http://www.imdb.com/help/show_leaf?votes