In your league, (how) do you determine "most improved" ? (1 Viewer)

HiveKueen

Two Pair
Joined
Feb 13, 2024
Messages
426
Reaction score
390
Rewards
208
Location
Rochester, NY
League Runners,
Do you try to have a "most improved" award in your league? Is so, how do you make the determination.

This will be my next project with chatGPT (aka Chatty).

Here's my start with this conversation (warning, AI babble coming ;)):

My kickoff question was:
Let's pivot a bit. what metric would be best for basing a most improved player on. I'm thinking avg pts per game but am wondering if there is an even better one?
Chatty's chatty response was...

This is an excellent pivot, because “Most Improved Player” is one of the hardest awards to get right — and one of the easiest to make feel unfair if the metric is wrong.

Let’s walk this carefully and land on something that is:
  • poker-realistic
  • statistically defensible
  • intuitive to players
  • resistant to gaming
  • aligned with your existing points system

Why “Avg Points per Game” is a good starting instinct​

You’re right to start there.

AvgPoints already:
  • normalizes for number of games
  • incorporates finish position, field size, pot, and cost
  • is familiar to players (they see points every game)
But by itself, raw ΔAvgPoints has two problems:

  1. Small-sample inflation
    A player who played 3 games one season and 6 the next can swing wildly.
  2. Regression artifacts
    Someone who had an unusually bad season can “improve” just by regressing to normal.
So AvgPoints is the right signal, but needs guardrails.


The core question you’re really asking​

“How do we measure genuine skill improvement, not variance or attendance?”
That means the metric must:

  • compare like with like (season to season)
  • discount tiny samples
  • reward sustained change, not one hot night

The best overall metric (recommended)​

MostIMprovedFormula.webp


Where:
  • AvgPts = average league points per game in a season
  • GP = games played in that season
  • S1 → S2 are consecutive seasons



Why this works so well​


1) Uses​


So it already captures:
  • field difficulty
  • buy-in discipline
  • finish quality

No new abstractions.

2) Square-root weighting controls noise​


Using:

min⁡(GPS1,GPS2)\sqrt{\min(GP_{S1}, GP_{S2})}min(GPS1,GPS2)
means:
  • improvement over 8→10 games matters more than 2→4
  • but volume still has diminishing returns
  • both seasons must have “evidence”

This avoids:
  • rewarding someone who barely played before
  • punishing someone who missed a few weeks



3) Directionally fair​

  • Improvement must be sustained across games
  • A one-night heater won’t dominate
  • Consistent, real growth wins



Why not just total points increase?​


❌ Total points is attendance-biased
❌ Encourages grinding, not improvement
❌ Penalizes players who play fewer games




Why not finish-position improvement?​


❌ Too noisy
❌ Field-size dependent
❌ Doesn’t reflect buy-in discipline or pot dynamics




Why not efficiency-only metrics?​


E.g. points per buy-in improvement

These are great secondary indicators but:
  • they miss survival skill
  • they can reward overly tight play
  • they’re less intuitive to explain


What I would not recommend​

  • Any metric using raw ranking change
  • Metrics that compare non-consecutive seasons
  • Metrics without minimum participation rules
  • Metrics based on single-game highs



How to explain it to players (important)​


One-sentence explanation:


“Most Improved Player is awarded to the player whose average performance improved the most from last season to this one, adjusted so that improvement over more games counts more than small-sample swings.”

That will land well.




Bottom line​


Best metric:
✅ Change in AvgPoints
✅ Weighted by shared games played
✅ Season-to-season only


It’s:
  • fair
  • poker-realistic
  • hard to game
  • fully supported by your existing data

If you want next, I can:
  • compute this award for all seasons
  • show how different formulas change the winner
  • help you choose the minimum-games thresholds

This is exactly the kind of award your league is ready for.
---end Chatty's response
 
And I thought running my 40-man golf league was a pain in the ass...
I'm retired, have no kids or g-kids, and this is my main hobby - i have about 6 league related projects going right now, in addition to running the games and planning for the championships. I should get a life.

Actually I'm also overseeing the complete remodel of a seasonal camp in the 1000 Islands, doing EOY tax planning and cleaning my house for a family brunch on Friday.

Ain't retirement grand!
HK
 
'excellent pivot!'
'you're right to start here!'
'this is exactly the kind of award your league is ready for'

Maaann, Chatty is laying it on THICK!

What does this mean?
❌ Doesn’t reflect buy-in discipline or pot dynamics
 
What does this mean?
❌ Doesn’t reflect buy-in discipline or pot dynamics
Ok...I asked Chatty what she meant by this . here's her answer...

3. “Doesn’t reflect buy-in discipline or pot dynamics”​


This one is subtle but important for your setup.


In your league:
  • players can rebuy
  • pot size varies
  • total cost varies per player
Finish position ignores all of that.


Two players can both finish 6th, but:
  • one did it on a single buy-in
  • the other fired 3 bullets
  • one played in a $660 pot
  • the other in a $240 pot

From a league-performance standpoint, those are not equivalent outcomes.


Your points formula explicitly rewards:


  • deeper survival in larger fields
  • doing so efficiently (cost-adjusted)
  • meaningful pot participation

Finish position throws all that information away.

She is right!
 
Last edited:
If the group appreciates it and places social value on it....go for it.

Seems like a bit of admin (on top of the admin that is unavoidable). In our group we just say that if you won the Championship trophy after having not won it in past seasons....good on you. And we consider that a reflection of improvement.
 

Create an account or login to comment

You must be a member in order to leave a comment

Create account

Create an account and join our community. It's easy!

Log in

Already have an account? Log in here.

Back
Top Bottom