Now that the NBA regular season has come to a close, I wanted to do a quick comparison of preseason win predictions as compared to the actual results. The sources used were:
How the sources compared to actual win predictions is below. The red-line signifies how much the actual wins were (i.e. Lakers finished at 17 wins, but ESPN had them at 27, which is why ESPN overrated them by 10 wins).
- Everybody overrated Houston, New Orleans, Phoenix & Philly
- Underrated was Toronto, Golden State & Portland
- Charlotte, Detroit, Miami and Milwaukee were more divisive
- The WS projections did the best in catching the sleepers like Portland & Charlotte, but it’s worth noting that they overrated a lot of teams, since they over-projected wins (average wins was ~43 when it should be 41).
- 538 performed best against Vegas at 17/30 correct. ESPN was second with 16/30, and SI was last at 13/30
If you assume you need 55% to break even with Vegas, 538 barely makes you a profit at a 56.6% success rate, and all the other sources lost you money. Looking closer at the data, a lot of the bets that 538 and SI WinShares lost were ones where their projections closely matched Vegas’s. Dallas for example, 538 had them at 38 wins, Vegas had them at 38.5 and they finished at 42.
So if you only made the bets where 538’s projections differed from Vegas’s by 3+ wins, it’s a 75% success rate (12/16 bets won), If you go where they differed by 5+ wins, it’s a 80% success rate (8/10 bets won),
This was just a quick exercise to see how different projection methods matched up to real life results — I often don’t go back and check the accuracy of preseason projections, so it was interesting to see how the “experts” matched up this year.