Youâre just making up words now.
- Iâm simplifying a bit, but Iâd say that the X-axis is basically the Average Net Efficiency for a team at a given point in the season. The Y-axis is the scoring metric that Torvik/KP use - you can also think of it as expected winning percentage. Thinking about this in time steps, if Team Aâs Average Net Efficiency is 0, this would put Team A directly at the midpoint of the curve on that bottom graph (since the scale on the X-axis of that graph is the ratio of points scored to points conceded, a 0 Net Efficiency is the same as 1:1 ratio between points allowed vs points scored - also known as the only time that 0 = 1). Now, letâs suppose that Team A goes out and has a +30 Net Efficiency in its next game. This will pull their Average Net Efficiency to the right, putting them at a higher point on the Y-axis. For Team A, and teams in a similar position, it doesnât take much movement in terms of Average Net Efficiency to dramatically increase their expected winning percentage, and thus ranking in Torvik/KP-world. If youâre a Purdue, then youâre already pretty far to the right of that curve where itâs the flattest, so itâs going to take a lot more to affect your overall score/ranking.
- Not really, (this is pure conjecture) but for both, I think that itâs less about how your actual performance compares to whatever the Vegas line is or whatever game result Torvik or KP predicts for you*, and more about how your Net Efficiency from that game affects your Average Net Efficiency. Since averages are very sensitive to outliers, Iâd expect blowout wins/losses to generally translate to significant jumps/drops in the both NET and Torvik/KP rankings.
*I have no mathematical proof that this is the case, but intuitively, I expect individual game impacts to your ranking to be heavily correlated with over or under-performing expectations, mainly because everything is adjusted for opponent.
The nerdery has only just begun, Hooâs ready to talk about Miyaâs Bayesian approach to evaluating/modeling college b-ball?
Itâs in the topic title, so you have no one to blame but yourself - 0 sympathy.
Love Evan Miya
Speaking of Evan Miya, I heard on 3 man weave podcast that heâs doing a new metric that cuts off or discounts blowouts. Prepare for a showdown with Kenpom over this, I guess
Ken is actually anti-margin-of-victory specifically when it comes to use in tournament selection, from his newsletter on what the WAC is trying with their conference tournament seeding:
The WAC system is better than the system used by any other league, and itâs also better system than the NCAA uses for its tournament, since the selection committeeâs rewards for wins are unknown. You know that Gonzagaâs win over USC helped them, but you donât know how much. On Selection Sunday, even committee members wonât be able to say, because itâs all subjective and impossibly complicated for humans to analyze.
Hope springs eternal that the NCAA will someday use a system like this. So far, theyâve shown no willingness to do so. Even where they had the opportunity to test something innovative in the NIT, they chose to not even try, and will simply reward teams based on NET ranking. The NET, previously described a sorting tool, now is the de-facto ranking method for the top-six leagues to earn a bid to the NIT. (Itâs very bad to have a ranking so heavily dependent on scoring margin to be used in this way.)
By contrast, the WAC Resume Seeding System gives the same reward to a team regardless of margin of victory. Itâs relatively simple, elegant, and transparent. And I hope itâs eventually used throughout college basketball.
The full newsletter is a good read for nerds: A salute to the WAC Resume Seeding System - by Ken Pomeroy
Well if Hartford is in there somewhere, this viz is wrong; theyâre D3 now.
Speaking of Evan Miya, I heard on 3 man weave podcast that heâs doing a new metric that cuts off or discounts blowouts. Prepare for a showdown with Kenpom over this, I guess
Iâm having trouble finding it now, but I remember reading on his site last year that he discards anything that happens after a teamâs probability of winning is something like 99.9999% (not the exact figure but basically once the outcome is assured). I wonder if heâs tweaking the garbage time definition or something.
@AdventiveQuasar were you the one advocating that they use Wins Above Bubble as the Tournament Selection method? I quite like that metric - it factors in the efficiency considerations while ensuring that winning games is what gets you into the Tournament rather than âstyleâ points.
Yep, thatâs one of my hobby-horses. The NCAA should do it, and free the committee from the prison of their making! Those people all already have jobs anyways!
Well, consider me converted.
Metrics aside, the NCAA hasnât even decided what their philosophy is for picking teams for the NCAAT in the first place. Do they want the best teams? The teams with the best wins? The teams with the best resumes? The teams with the toughest schedules? Until they decide that, no one can create an appropriate ranking tool, because they donât know what theyâre ranking. In fact, I think the NCAA views it as a feature, not a bug, that no one knows exactly what the NET is measuring: âWe donât know what we want, and we donât know what NET measures, so itâs a perfect sorting tool!â
One of the biggest selling points of WAB (for me) is simply that it has a clear underlying philosophy. âThis is what we value, and this metric measures it.â
They have a bit of a marketing problem. If theyâre too close to normal efficiency ratings / 4 factors stuff, whatâs the point? But if theyâre too far away, and look wacky, then youâre in âLOL, you crazy!â territory.
Moving the Shot Quality discourse to the nerd thread. I agree, I wonder if they reacted some to getting trolled a bunch on social media when they would talk about a âshot qualityâ win for a team that lost. I think thatâs a marketing/framing issue though; soccer has a similar metric called Expected Goals (how many goals a team would be expected to score off the profile of the shots they take), and certain segments of fans absolutely hate it.
Ha ha! Yeah, I definitely trolled them some on here after one of the tourney games last year. Some team had a SQ win in a big upset. Canât remember the detailsâŚ
I wonder if they reacted some to getting trolled a bunch on social media when they would talk about a âshot qualityâ win for a team that lost
But isnât that sort of situation essentially the whole point of measuring Shot Quality? I donât get why people would complain about that.
On individual defensive ratings: individual defensive rating is basically team defensive rating while you are on the court +/- individual steals, blocks, and defensive rebounds. So it will be worse for anyone who doesnât get steals, blocks, and defensive rebounds.
Thereâs an assumption in the calculation that everyone is equally good at forcing misses from opponents: Calculating Individual Offensive and Defensive Ratings | Basketball-Reference.com
- Out of necessity (owing to a lack of defensive data in the basic boxscore), individual Defensive Ratings are heavily influenced by the teamâs defensive efficiency. They assume that all teammates are equally good (per minute) at forcing non-steal turnovers and non-block misses, as well as assuming that all teammates face the same number of total possessions per minute.
That is one reason individual defensive ratings and on-offs tell different stories about some guys (Rohde and IMac most notably). If you are a good positional defender but donât force steals, get blocks, or get rebounds, the individual defensive rating formula assumes you are worse on average than your teammates. Assigning individual credit for a team stat is complicated and prone to error, especially when the box-score does not have all the relevant information.
For reference, teamâs on-offs at a glance: https://hoop-explorer.com/TeamReport?baseQuery=&filter=&gender=Men&incRapm=false&incRepOnOff=false&maxRank=400&minRank=0&showOnOff=true&sortBy=desc%3Aoff_poss%3Aon&team=Virginia&year=2023%2F24&
Feels like thereâd be room for a decent defensive stat thatâs more labor intensive**. I suspect teams do have stuff like that. But I also suspect they have more what Iâd call âdataâ and would be better off partnering with a stats nerd to stat-ify the data, if that makes sense
** like, how often was a guy targeted. Howâd he do? Etc