2016: The FP System

Back in 2014, I wrote an introduction to the primary “advanced” stat I use to evaluate prospects – FP. Since then, I’ve refined the way I use it to project, but the base stat has stayed the same. FP stands for fantasy points – it’s really nothing more than a way to convert basic statistics in to fantasy points for fantasy college basketball. However, that means it’s also a general effectiveness metric – somebody who gets stats is typically contributing. When I started applying it to the draft and NBA players, I found that FP did a decent job identifying players who were most likely to become superstars. More importantly, it did an even better job identifying overrated players. Knowing who not to draft is just as important as knowing who to draft.

FP can be calculated as follows:

1 FP per point
1 FP per rebound
2 FP per assist
3 FP per steal
3 FP per block
-2 FP per turnover

This provides a nice baseline to judge players. I typically then reduce it to FP per minute (FPm). It’s an easy if imperfect way to quickly judge players. When it comes time to talk about individual players and create my big board, I weigh other factors, but for a week to week quick look at guys, FPm does the job well. Is it less “advanced” than other advanced stats people use for the draft? Yes. But that is not the question to ask. Is it better than other advanced stats at identifying good and bad prospects? Um…I’ve only been doing big boards for two years now, which means it’s a bit premature to say, and because it’s a system that looks to fit current data to historical trends, it will look better than average against history. That being said, I think it’s at least equal to other advanced projection systems out there. Otherwise, I wouldn’t use it.

Last year, I added in age, level of competition, and other modifiers. Looking back, I think it made the system less accurate, not more. Why? Well, turns out there’s something to Occam’s Razor. It’s really tempting to tinker with something until it’s perfect, but the more complicated something becomes, the easier it becomes to lose sight of what made it successful in the first place. And that happened. So what exactly is FP good for, and what’s the best way to use it?

Well, here’s the thing. There is not a system out there that will be perfect. No statistical system, no scouting system, no hybrid. So what, then, is the purpose of even developing a system? Ultimately, the draft, just like basketball itself, is a game of percentages. There’s no such thing as a 100% can’t miss prospect. Injuries happen. Lack of development happens. Bad team fit happens. Shit happens. The goal then is not to hit on every pick, but to hit at as high a percentage as possible. And it would do everybody well to keep in mind to think of prospects on two axes: upside and likelihood.

The most common issue I see in typical prospect evaluation is overprojecting everybody. Comparisons, rather than picking the most likely outcome, often identify something closer to a 90th percentile comparison. Almost all players will fail to meet this. While it’s great for hyping prospects and generating interest in the draft and in rookies, it’s actually really terrible for creating realistic expectations. It also leads people to completely disregard the “likelihood” axis, and sometimes even the “upside” axis. If players were projected to their most likely outcome rather than to some subjective standard, most players would generate very little interest. It’s the simple truth. Most prospects fail.

FP tries to identify players who rank highly on both axes. Understand that, by highly, we’re still often talking about less than a 25% chance of actually becoming a useful NBA player. For truly top prospects, it’s a higher chance of becoming a useful NBA player, but still low chance of hitting their ceiling. Therefore, it also identifies players who rank very low on chance of actually becoming a useful NBA player. And ultimately, while you may be talking about a 10-15% difference between a mid-level prospect and a non-prospect, teams that consistently go after the right prospects will eventually get clear on-court results.

So, what are the biggest indicators for production at the next level? FP per game over 30. FP per minute over .9. Those are the magic numbers. Basically, guys who are productive in college tend to be more productive in the pros than guys who are not. People often get hung up on what tools a player has, but while tools matter, being able to use tools matters just as much, if not more. A hammer is only useful with a nail if you can actually hit the nail. But even productive college players typically don’t become productive NBA players. Why? Well, we’re back to the fact that most guys fail. You can identify 25+ guys who might succeed, but only a few of them will. What we’re left with are the questions everybody is still trying to answer: is there a way to better identify players who have a higher than typically expected chance to become a useful NBA player? Is there a way to better identify players who have a lower than typically expected chance to become a useful NBA player?

If you’ve been following this draft series, you may see where this is going: the very first split is NBA role/fit. Players who don’t have a clear NBA role or fit are much less likely to succeed. This includes non-primary ballhandlers who can’t play defense, non-big non-ballhandlers who can’t shoot from 3, bigs who can’t defend the rim, bigs who are poor in the offensive pick and roll/pop, etc. A lot of prospects can be dismissed just based on these criteria. There’s a lot of college “bigs” who simply aren’t big or athletic enough to succeed at the next level. There’s a lot of college guards and forwards who aren’t good enough shooters or handlers to succeed at the next level.

After the players who don’t fit in the NBA are separated out, then age becomes a factor. Far too many people start with age first. There are two major reasons why starting with age instead of production is a bad idea. First, production is simply a better predictor of future performance than age. Second, it assumes that a younger player will always develop, and develop to be better than older players. This is a bad assumption. Many players remain flat or show little improvement, and some even prove to be worse as they get older. This being said, a young productive prospect typically has higher upside, higher likelihood, or both compared to an equally productive older prospect, so age should certainly be factored in. But productive older players are pretty much always underdrafted and tend to be the biggest draft misses year after year. After the great freshmen, great non-freshmen should always be ranked ahead of the simply good freshmen. Talent just matters more than age.

Then there are a couple of other factors worth considering but which don’t particularly move the needle much for me, at least after factoring out players who don’t have clear role. Level of competition matters, but most low and mid major prospects get split out at the role/fit level, and those who aren’t are worth keeping an eye on. As Steph Curry and Damian Lillard show, discard low majors at your own peril. Physical tools are worth noting, but as said above, tools that don’t manifest in on-court production aren’t particularly useful. That being said, better tools do indicate slightly higher upside. I also take a look at things like college team pace, college team quality, competition level splits, and a few other little things, but only for purposes of splitting players up within tiers.

Ah, right, tiers. Ultimately, as mentioned before, I still use a subjective component when ranking prospects. I use FP to set players into tiers, but sorting players within tiers…well, it doesn’t even need to be done, but people like full rankings. For splitting hairs, I use subjective judgment. It’s just impossible to get too fine with data like this. Because there’s two axes to consider, how do you value high-risk, high-reward prospects compared to low-risk, low-reward prospects? How do you compare two players separated by just a little bit when you’re dealing with a small sample and two or three more games could see them flipped? The best way is just to put them in groupings that indicate generally how valuable a prospect is. So that’s what I do. My rankings within tiers are far less important than which tier a player is placed in.

So, that’s my system. It’s a system designed not to be perfect, but to take a broad-level look at prospects to try to just play the odds. To date, it’s hit enough in both directions that I feel comfortable relying on it as a prospect evaluation tool. There’s plenty of other tools out there. Many of them have different goals with different pros and cons. When thinking about evaluating prospects, be aware of how the evaluator is trying to evaluate. And if you don’t know…you may want to reconsider using that source as a reliable source. Always ask how/why, and if it doesn’t make sense, well, move on. There’s certainly no dearth of analysts. This is how I analyze.