Improved Ranking System ???


#141

Yeah, my SQL sucks, but I know enough to know it’s a lot more complicated :grinning:.


#142

Can tell you right now that it’ll give way more weight to Pinburgh than IFPA due to how many of the top 250 are in the field.


#143

The implementation isn’t too difficult, it’s the testing, tweaking, retesting part and that takes alot of time. If it is something that affects the entire system going back to the first tournament, it takes forever to run. :frowning:

Not to mention the pay sucks


#144

Get back in your cage! You are NOT allowed to speak!

Next thing you know @Vengeance will start bitching about his pay :slight_smile:


#145

Strongly agree that it’s way too early for putting on the IFPA site. Sensitivity testing required. My sql isn’t up to snuff either, but I can do some ad hoc testing.


#146

Now we just need to remove league stats / points from the qualifying of SCS / ACS and we’d be good!


#147

[/quote]
Done: https://www.ifpapinball.com/rankings/power100.php?type=nl
[/quote]

I made the list. :slight_smile:


#148

An unfair advantage- or a system doing what it’s intended to do… encourage participation!

People keep forgetting- if you aren’t competing you aren’t succeeding in the rankings… it should be no surprise by now, that is by design!


#149

Some interesting observations when you get inside it.

For some players just a few large events, like Pinburgh, can drive their results. I found one player where over 1/3 of their “vs” is from one Pinburgh. In those situations, you can move way up or down in a hurry - - such players’ positions are unstable. Also, their power rankings can be heavily driven by how well they did in those one or two events, i.e. driven by one or two good or bad weekends. Such players’ power rankings as thus more suspect IMO.

For most players, the weighted-average rank of their versus is not 125.5 (the midpoint of 1-250), but a higher rank, usually between 90 and 100. This is because the higher-ranked players play more, so everyone’s versus stats will tend to skew towards higher ranked values. This is true of both “circuit regulars” and most medium-activity level players. Due to this activity level of the higher ranked players, very few players with lots of data points will have a meaningfully “harder” average versus rank than the rest. “Strength of Schedule” is a non-issue for at least 90% of the Power 100. This held true for both North American and European players.

Some players’ percentages are really slow to move. For instance, Trent has over 3300 data points, so no one event, not even Pinburgh, is going to move his percentage greatly.

For a fairly small fraction of the list, including but not limited to the “down under” ones, a high proportion of their total versus stats come from just a handful of players. In many of those cases, their versus players are not in the top 100 but in the 101-250 zone. Basically, players who don’t travel much and haven’t been to many [or any] large events like Pinburgh or EPC tend to skew towards weaker average opposition as measured by IFPA ranking.


#150

Those are interesting insights, thanks!

To deal with the situation where percentage essentially stops moving due to the large number of events played (as in Trent’s case), would it make sense to put some expiry algorithm on the percentage or, alternatively, add a second percentage hat has a decay function similar to the decay function for the ranking?

When I look at someone’s percentage rating, I’m probably not that much interested in having scores that are more than five or ten years old influence the value. What happened in the past three years or so is probably more relevant.


#151

The time period is the same for all players in the metric … 3 years (all ‘active’ events)

It just looks like 30 years worth of results for Trent :wink:


#152

Ah, I didn’t know that. So, the percentage of points gathered already decays the same as the rankings. That’s good to know, thanks!


#153

Some other stats / trivia, data updated to today:

Max Count - Trent, 3281; Johan Genberg #2 with 2906
Min Count - Peter Watt, 176
Average - 1272
Median - 1192

Top 30 count-wise accounted for 50.1% of all data points within the Power 100.

Win Percentage did not correlate with data point count.

Collective Win Pct of Power 100: 64.3%

Europeans are more likely to have ties: top 12 in portion of win-loss-tie record that are ties are all Europeans.
Highest Pct ties - Nicklas Lindahl, 8.9%
Lowest Pct Ties - Escher Lefkoff, 1.0% [also Cryss Stephens at 1.0%]

Pinburgh 2017 had 125 players ranked in the top 250 at the time as participants. That’s a lot of win-loss-tie points for everyone who played in it! For those who played in it each of the last 3 years, on average those 3 Pinburghs represent about 1/4 of their total record.


#154

I guess there is no easy way to see “last 12 month” power 100 ranking to see if there is improvement? :slight_smile:

Those early day PAPA Classics half ass tries sure hurts!


#155

Any further thoughts/responses since the Power100 going live?


#156

I’m thinking you’re walking around with your chest puffed out more than it used to be . . . but now you have an internal rage to kill Andrew Foster? :slight_smile:


#157

Hahahaha, oh Josh you know me so well, I’m at least 3 inches taller now :blush: Leaving my home is actually becoming a struggle, everyone is desperate for a peice of me! 'Oh my god it’s him…he’s on the ‘Best of the best’ list. It’s my cross to bear. It’s boring, but it’s part of my life :blush:

But seriously, is this going to be something that now just fades away into obscurity under the ‘Custom’ tab. Or does it have any merit to carry any weight? I think Per’s comment on the website is relevant

‘Great initiative! Have you considered having a minimum number of games to be counted? I see that Zach Sharpe has over 2200 games, while for example Peter Watt doesn’t have 200. I think the NBA does something similar when it comes to shooting percentage. I.e. you can’t just shoot and make 1 3-pointer to be counted’

And for good measure, seeing as I was baited into it, I’ll just leave this here https://www.ifpapinball.com/players/compare.php?p1=9560&p2=2522 (15-10 :wink:)


#158

Any weight towards . . . what? You said your only interest in WPPR’s now is qualifying for the WC. So I’m guessing you’re asking if we can somehow make this part of the IFPA WC qualifying process?

I haven’t given much thought to putting in a minimum in order for a player to be counted. At some point this metric is meant for the players that play less right (either by choice or because of their environment)?

So the average of this Power 100 list is 1300 matches . . . should Peter be included at 200 matches? Should you be included at 555 matches? I have no idea where that line should be, and I like the fact that this highlights the people that “Don’t travel to play, but seem to be a solidly skilled player”. Without you guys on this list, it’s just another shake up of what the other metrics are showing anyway.


#159

Is there anyway for the people outside the top 100 to see where they stand? Ty


#160

Nope . . . you’re either IN the POWER 100 . . . or you’re a scrub :stuck_out_tongue:

What you can do is take this URL and put in your player ID at the end:

https://www.ifpapinball.com/players/compare250.php?p=23396

That will pull up your POWER 100 stats, so @PinballProfile for example, you can see that his winning percentage is 43.11% against the top 250. Looks like the cutline to be in the top 100 is 51.93% at the moment.