|
Post by Glenn-Philadelphia on Jan 12, 2023 11:20:53 GMT -5
Ok so we had a small adjustment to the scoring algorithm which was separate from some other adjustment and the addition of the zone shot total grab that you put in place. Understood. I have 32 games of goal data vs several data pulls of different time periods worth of games that suggest there is a strong trend in what I am suggesting. I understand small sample sizes are going to have higher likelihoods of an outlier and stronger ones at that. But year in and year out our best goaltenders have been average performers. I've finally narrowed down the causes a bit, I've shown you time and again that there seems to be a correlation between goals in the zones being off particularly for goalies with strong cards and everything you have pulled seems to corroborate this. What else do you need to show that there is a calculation error going on somewhere? I'm not asking you to revamp your entire table. I'm asking you to ensure that the calculations being utilized are correct and/or that the table is being referenced correctly, especially when a goalie is supposed to have an advantage over the shooter. There is nothing above that shows that Shesterkin seems to be seeing an abnormal number of shots from Z14 that would suggest his 42% of goals from Z14 makes sense. You can track it as long as you want. I've already shown you the problem. I've spent two years dealing with this, first with Nedeljkovic last year having the what, third best card or something. And again this year with Shesterkin having the best goalie card. Both ended up being average at best. When was the last time our best goalies were at the top of our Save% and GAA leaderboards? I know we're not going to go back and rerun the season but it'd be nice if we could have the problem fixed so that my team is actually playing on a level playing field. Right now though it seems like we'll just keep playing with one hand tied behind Shesterkins back and watch my team once again be completed shafted. I'd ask for at least an apology if/when you finally diagnose and fix this since my team has been so heavily affected for two years in a row and an acknowledgement that my teams opportunities at being even halfway decent, that I've spent literal years working to build, are being destroyed by bad code. I mean its not like my team started on even footing to begin with but hey as long as I keep paying my dues who cares right? Below is the zone 14 scoring percent chances. The row used is based upon several factors with the primary factor being goalie stat chosen compared to shooters shoot % reduced to a 0 to 10 rating. Theses value are modified by how fresh the shooter is regarding his wind (stamina, length of current shift, time played during game) as well as a calculation for the goalies where they are always considered fresh (their shots faced or activity is not currently calculated and always is considered best). Other factors include if a team is on the PP or SH as well as possible random modifiers for deflections or screens. If the derived shooter shot % is 5 and the goalies derived stat is a 7 then the player would have a 15% chance to score on the shot (from the chart below). When viewing the chart below and applying the formula above you see that better goalies are not being punished for having better ratings. In fact, it's quite the contrary as you see that the poorer valued goalies have a larger percentage of increased scoring chance when the comparison of shooter to goalie enters the +3 row and higher. Zone 14 MOD PCT -9: 2 -8: 4 -7: 6 -6: 8 -5: 10 -4: 11 -3: 13 -2: 15 -1: 17 +0: 19 +1: 20 +2: 22 +3: 25 +4: 28 +5: 31 +6: 34 +7: 38 +8: 42 +9: 49 The engine is complex and trying to reduce the concept/function of any one thing in particular is much easier in a live interaction. I would be glad to talk to anyone if they have specific questions or would like more detail on something. This could be a one on one zoom/phone call or a zoom call if multiple owners want to find out more or have their questions answered as well. That offer remains open.
|
|
|
Post by Chris-Suffolk on Jan 12, 2023 11:50:05 GMT -5
Glenn, Reading the above statement regarding: "Theses value are modified by how fresh the shooter is regarding his wind (stamina, length of current shift, time played during game) as well as a calculation for the goalies where they are always considered fresh (their shots faced or activity is not currently calculated and always is considered best)."
Isn't that giving an Advantage to the goalie?
I imagine it must be insanely difficult to give the goalies a deduction to stamina in game, but if its being done to skaters...that's unfair
|
|
|
Post by Glenn-Philadelphia on Jan 12, 2023 14:52:29 GMT -5
Glenn, Reading the above statement regarding: " Theses value are modified by how fresh the shooter is regarding his wind (stamina, length of current shift, time played during game) as well as a calculation for the goalies where they are always considered fresh (their shots faced or activity is not currently calculated and always is considered best)." Isn't that giving an Advantage to the goalie? I imagine it must be insanely difficult to give the goalies a deduction to stamina in game, but if its being done to skaters...that's unfair Well, yes and no. Goalies play 100% of the game, forwards 25-33% and defenders 33 to 50%. In the past I had tossed some ideas around in my head where goalies could tire after a flurry of shots and recuperate accordingly after having some period of normal or lower than normal shots. When dealing with the goalies stamina, they are handled outside of the in game wind via their color going into the game. A blue goalie will have a few extra points spread across his stats, a green goalie will have no adjustments and yellow, orange and red goalies will have mild to significant reductions to some of their scores as they progress through these colors. This may be revisited in the future but only after GHL scoring stabilizes.
|
|
|
Post by Chris-Suffolk on Jan 12, 2023 15:00:16 GMT -5
Glenn, Reading the above statement regarding: " Theses value are modified by how fresh the shooter is regarding his wind (stamina, length of current shift, time played during game) as well as a calculation for the goalies where they are always considered fresh (their shots faced or activity is not currently calculated and always is considered best)." Isn't that giving an Advantage to the goalie? I imagine it must be insanely difficult to give the goalies a deduction to stamina in game, but if its being done to skaters...that's unfair Well, yes and no. Goalies play 100% of the game, forwards 25-33% and defenders 33 to 50%. In the past I had tossed some ideas around in my head where goalies could tire after a flurry of shots and recuperate accordingly after having some period of normal or lower than normal shots. When dealing with the goalies stamina, they are handled outside of the in game wind via their color going into the game. A blue goalie will have a few extra points spread across his stats, a green goalie will have no adjustments and yellow, orange and red goalies will have mild to significant reductions to some of their scores as they progress through these colors. This may be revisited in the future but only after GHL scoring stabilizes. Fair enough for the goalies. I'm good with that explanation. What type of value deductions are the forwards/defenders receiving in the above?
|
|
|
Post by Glenn-Philadelphia on Jan 12, 2023 16:04:03 GMT -5
Well, yes and no. Goalies play 100% of the game, forwards 25-33% and defenders 33 to 50%. In the past I had tossed some ideas around in my head where goalies could tire after a flurry of shots and recuperate accordingly after having some period of normal or lower than normal shots. When dealing with the goalies stamina, they are handled outside of the in game wind via their color going into the game. A blue goalie will have a few extra points spread across his stats, a green goalie will have no adjustments and yellow, orange and red goalies will have mild to significant reductions to some of their scores as they progress through these colors. This may be revisited in the future but only after GHL scoring stabilizes. Fair enough for the goalies. I'm good with that explanation. What type of value deductions are the forwards/defenders receiving in the above? Of the top of my head: Playing out of position if not on PP or SH Fatigue (the more tired the less chance of maximum stat value) Home Ice Advantage sometimes adjusts by 1 (determined when action is resolved) Fear Factor can potentially play a factor (more so for significant differences in team toughness ratings)
|
|
|
Post by Dane-Hamilton on Jan 12, 2023 16:58:32 GMT -5
Wait isn't that backwards?
Just went and looked at the gaps. The minus categories (where goalies have advantage) have gaps of 2% between and 1% between -5 and -4.
However when goalies at a disadvantage (positive numbers) the shooters gap between is 3% per point they get on the goalie except 4% between 5-8 and 7% between 8-9.
Unless I'm reading it wrong. So if a shooter + 5 they have a 12% bonus to base (0) but if a Goalie has -5 shooter only hurt by 9%.
|
|
|
Post by Jon-Seattle on Jan 12, 2023 17:31:39 GMT -5
Ah I think I see what you’re saying Dane. Seems like it’s just more of a disadvantage to have a bad goalie but I see your thought process.
I think the issue I have isn’t the calculation itself but more if it’s actually being implemented and calculated right as based on this Shesterkin should have at least an .800 save% in zone 14. And that’s only if he was facing shots from the top 10 shooters in the league at their absolute best at all times. I think he’s been significantly worse than that and my hunch is that the engine is taking the absolute values (rather than the - value) in this particular zone for goalies like Shesterkin
|
|
|
Post by Dane-Hamilton on Jan 12, 2023 18:19:47 GMT -5
How did you get that he must have %.800 in zone 14. Theoretically he could be shit at dice roll and get scored on every shot in a game, better odds of struck by lightning but possible.
Also Glenn is it a dice roll to see what # is used in that derived stat?
Example.
Shesterkin Glove - 8 vs Pavelski - 6
Doesn't Shesterkin roll essentially an 8 sided die and Pavelski rolls a 6 sided die. Pavelski could roll a 6 and Shesterkin could roll a 1 and Pavelski gets a big advantage that shot.
I could be misremembering though.
|
|
|
Post by Dane-Hamilton on Jan 12, 2023 18:23:49 GMT -5
Also just thinking if it is a dice roll like the example above maybe limiting minimum across the board would be helpful. I would say something like can't drop more than 5 from card stat value.
Using same example as above
Shesterkin could only roll a 3-8, but Pavelski can still roll 1-6. I don't know just spit balling
|
|
|
Post by Glenn-Philadelphia on Jan 12, 2023 19:31:51 GMT -5
The base off vs def calc is as follows. Off score - Def score. So, if a shooter ends up with a 5 and the goalie ends up with a 7 the final calculation is 5-7 or -2 meaning the -2 column/row is used. This table holds the chance to score not the chance to save percentage.
If Shesterkin has a glove of 8 then because he is fresh the sim will in effect roll 3 8 sided dice (3d8) for him and take the best value. If the number come up 2,3,7 he would get a 7. If the numbers come up 2,3,3 he would get a 3.
Skaters use the same if they are Fresh. Based on their current wind their is a stepped down approach something like this:
2d8 and take highest 1d8 2d8 and take lowest 3d8 and take lowest
I would have to check the code but those are what the value are I am 99% certain.
Again, given the current state, and the state of the engine for 10+ years, goalies always get the 3dvalue and take highest.
|
|
|
Post by Glenn-Philadelphia on Jan 23, 2023 15:44:46 GMT -5
187 GP (games through 1/22/2023). I am going to do some more macro review on stats such as goalie save pct and shots on goal per game and will post those when done.
SOG Pct (of all shots from zone, how many were on net) Z15: 53.6% Z14: 61.2% Z13: 54.8% Z12: 42.3% Z11: 51.2% Z10: 42.7% Z09: 33.0% Z08: 39.3% Z07: 33.7% Z06: 27.7% Z05: 32.2% Z04: 29.2% Z03: 24.9% Z02: 26.9% Z01: 22.2% Goalie Save Pct (Save pct by zone) Z15: 0.936 Z14: 0.82 Z13: 0.933 Z12: 0.902 Z11: 0.868 Z10: 0.91 Z09: 0.922 Z08: 0.919 Z07: 0.938 Z06: 0.944 Z05: 0.922 Z04: 0.938 Z03: 0.986 Z02: 0.987 Z01: 0.994 Shots on Goal (Total shots on goal by zone) Z15: 1754 Z14: 2268 Z13: 1870 Z12: 1088 Z11: 1887 Z10: 1109 Z09: 667 Z08: 926 Z07: 725 Z06: 682 Z05: 848 Z04: 762 Z03: 144 Z02: 158 Z01: 156 TOT: 15044 SOG % by Zone (from table above, % of shots by zone) Z15: 11.7% Z14: 15.1% Z13: 12.4% Z12: 7.2% Z11: 12.5% Z10: 7.4% Z09: 4.4% Z08: 6.2% Z07: 4.8% Z06: 4.5% Z05: 5.6% Z04: 5.1% Z03: 1.0% Z02: 1.1% Z01: 1.0%
Teams are averaging 40.2 Shots on goal during this time period
|
|
|
Post by Jon-Seattle on Jan 24, 2023 11:37:09 GMT -5
Latest Shesterkin game: 3 goals against all from Z14 on 41 total shots.
Since my last post he has started 5 games and allowed 7 goals. 6 out of those 7 goals were from zone 14.
His overall save % has bounced up due to a SO but the Z14 numbers aren’t getting any better and with the dice rolling mechanic we use all statistics should regress towards the average over time as the number of rolls drastically increases. (Which is why Danes suggesting of Shesterkin just being bad at dice seems unfounded)
|
|
|
Post by Scott-New York on Jan 25, 2023 16:17:27 GMT -5
...or, he sucks from zone 14
|
|
|
Post by Dane-Hamilton on Jan 25, 2023 17:33:52 GMT -5
Overall these look pretty good, that website with NHL shot data actually has our zone 14 as 22% of total shots. I'll have to look at goal % from there too if they have it.
|
|
|
Post by Chris-Suffolk on Jan 25, 2023 19:52:44 GMT -5
I'm confused! Are we comparing ourselves to the NHL or Not?
|
|