I need to spend some more time with the article, but my first impression is that the RISK definition isn't right. In the NBA I think that a team's RISK is that players will underperform their expected value. Conversely, REWARD is when a player over-performs his expected value.
The composite of Risk and Reward should correlate with team winning. However instead of a pure per game stat, I think that the players total results need to be factored in. For example, Joel Embiid was 6th in WS/48 min (BB REF) and Tyrese Maxey was 95th. In total WS Maxey was 24th and Embiid was 28th. Availability matters too.A blend of Total and Per Game stats helps screen out performances that aren't likely scaleable. A minimum minutes criteria could also help with screening. RW III played only 6 games and MitchRob played less than 800 minutes total.
Anyway, just my initial thoughts, thanks again for the article.
In our approach with MPT, the expected value and 'reward' are based on historical performance. We're not specifically looking for outperformers, but instead, the model focuses on players who can consistently "play their role" and deliver the type of returns expected of them. This approach isn't about finding undervalued players who might outperform but ensuring consistent contributions to team success.
In my opinion, our definition of risk aligns with this objective. By using the standard deviation of a player's return, we're identifying consistent performers who can reliably deliver their expected returns season after season, minimizing the risk of underperformance i.e. lower standard deviation means lower risk of underperforming against historical averages.
We have applied a minimum minutes criteria on a per game basis but you are right, availibility matters. To improve the model we could factor in player availability (e.g. discount for injury prone players) similar to adjusting for liquidity risks in financial markets for exemple.
Thanks again for the feedback and let me know what you think!
Thanks, very thought provoking.
I need to spend some more time with the article, but my first impression is that the RISK definition isn't right. In the NBA I think that a team's RISK is that players will underperform their expected value. Conversely, REWARD is when a player over-performs his expected value.
The composite of Risk and Reward should correlate with team winning. However instead of a pure per game stat, I think that the players total results need to be factored in. For example, Joel Embiid was 6th in WS/48 min (BB REF) and Tyrese Maxey was 95th. In total WS Maxey was 24th and Embiid was 28th. Availability matters too.A blend of Total and Per Game stats helps screen out performances that aren't likely scaleable. A minimum minutes criteria could also help with screening. RW III played only 6 games and MitchRob played less than 800 minutes total.
Anyway, just my initial thoughts, thanks again for the article.
Thanks for your insights - much appreciated.
In our approach with MPT, the expected value and 'reward' are based on historical performance. We're not specifically looking for outperformers, but instead, the model focuses on players who can consistently "play their role" and deliver the type of returns expected of them. This approach isn't about finding undervalued players who might outperform but ensuring consistent contributions to team success.
In my opinion, our definition of risk aligns with this objective. By using the standard deviation of a player's return, we're identifying consistent performers who can reliably deliver their expected returns season after season, minimizing the risk of underperformance i.e. lower standard deviation means lower risk of underperforming against historical averages.
We have applied a minimum minutes criteria on a per game basis but you are right, availibility matters. To improve the model we could factor in player availability (e.g. discount for injury prone players) similar to adjusting for liquidity risks in financial markets for exemple.
Thanks again for the feedback and let me know what you think!