Artifootball Intelligence: using artificial intelligence to find the best redraft strategies (Fantasy Football)

As the use of AI explodes around the world, I started thinking about how to apply its amazing capabilities to fantasy football. One of the easiest ways is to use it to draft robots. We’ve seen this simple application of AI in fantasy football for many years through auto-drafting, where players can give up their fantasy drafts and allow the computer to make player selections for them.

Similarly, I programmed 10 AI draft bots, all with different strategies, and simulated 1,000 drafts (randomizing the order each time) to determine which draft strategies are most successful. The main difference between each bot is when they select their first TE and QB. This project was done in preparation for the 2023 season, some of the results will be more specific to current NFL talent.

Let’s take a look at the robots! For reference, the MVP is determined by the player who accumulated the most points per game in the 2022 season.

AI bot strategies

AI Bot one:

• Draft the first QB in Round 4
• Draft the first TE in Round 6
• Draft the best non QB/TE otherwise
• Draft the best player ever in rounds 13-15

AI Bot Two:

• Draft the first QB in Round 5
• Draft the first TE in Round 6
• Draft the best non QB/TE otherwise
• Draft the best player ever in rounds 13-15

AI Bot three:

• Draft the first QB in Round 4
• Draft the first TE in Round 9
• Best available non QB/TE otherwise

AI Bots Four:

• Draft the first QB in Round 3
• Draft the first TE in Round 7
• Draft the second TE in Round 8
• Draft the best non QB/TE otherwise
• Draft the best player ever in rounds 13-15

AI Bot Five:

• Draft the first QB in Round 14
• Draft the first TE in Round 4
• Draft the best non QB/TE otherwise
• Draft the best player ever in rounds 13-15

AI Bots are:

• Draft the first QB in Round 8
• Draft the first TE in Round 4
• Draft the best non QB/TE otherwise
• Draft the best player ever in rounds 13-14
• Draft the best available QB in Round 15

AI Bot seven:

• Draft the first QB in Round 9
• Draft the first TE in Round 1
• Draft the best non QB/TE otherwise
• Draft the best player ever in rounds 13-14
• Draft the best available QB in Round 15

AI Bots eight:

• Draft the first QB in Round 5
• Draft the first TE in Round 9
• Draft the best non QB/TE otherwise
• Draft the best available QBs and TEs in Rounds 14 and 15

AI Bot Nine:

• Draft the first QB in Round 1
• Draft the first TE in Round 6
• Draft the best non QB/TE otherwise
• Draft the best available QBs and TEs in Rounds 14 and 15

AI Bot ten:

• Draft the first QB in Round 1
• Draft the first TE in Round 13
• Draft the best non QB/TE otherwise
• Draft the best available QBs and TEs in Rounds 14 and 15

The results

I’ve plotted the distributions of total starting points for each strategy below for Bots 1-5 and 6-10, respectively. Out of the box, Bots 1, 4, 6 and 9 stand out, averaging about 4-5 more points per game than rookies.

By bringing these four hallmarks together, we summarize what each of their strategies is.

• Robot AI 1 (Blue): QB early-mid shift and TE mid shift
• Robot AI 4 (Brown): QB at the start of the shift and TE at the middle of the shift
• AI Bot 6 (green): Mid-round QB and start-of-round TE
• AI Bot 9 (Red): QB in the first round and TE in the middle round

You will notice that the best performing strategies feature mostly intermediate QB and TE selections. While there are fewer strong options at the TE and QB positions, the gap between the top and bottom is much smaller than that at the RB and WR positions. This is why it’s generally important to load talent into those departments first.

However, you can still be successful by selecting a QB or TE early – the gap between strategies is quite small, especially at the top. As seen with Bot 9, the strategy of selecting a QB in the first round averages only three points less than the main strategy (Bot 4). The same can be said of picking a TE in the first round (Bot 7): this strategy averages only four points less than its owners’ best strategy! It’s important to note, however, that either strategy drawing a QB or TE picks the other first in a mid to late round. When I adjusted the AIs to draw both a QB and TE in the first two rounds, the starters’ total point average dipped below 100.

Finally, I plotted the points-per-game distribution of all players on each team in the draft—the results are wildly different. AI Bots 7, 8, 9, and 10 average over 10 points per game more from the entire team than the rest of the pack. What’s different about these four strategies? They all draw a TE or QB in the first round and/or prioritize the acquisition of backup QBs and TEs in subsequent rounds. This shows that these strategies may be stronger at producing deeper teams, and that a late-shift QB or TE tends to be more valuable than a late-shift RB or WR.

Conclusion

All in all, it’s smart save your QB and TE selection until mid-draft, from choices four to nine. However, while it’s typically smarter to hang on during redrafts to take these players, some QBs and TEs performed incredibly well last season, making their fantasy an unprecedented value. That means drafting them early could pay off if you can get the top guys (Mahomes, Allen, Hurts, Kelce). Be careful, however, as we saw that targeting both positions of power early on was detrimental in the simulations.