|
For Ongoing Reference
"Searching for Value" evaluates which money-line plays have the highest projected expected value in each week's NFL games. (I will also periodically look at NFL futures.) As sportshandle.com summarizes, "expected value... measures the probability gap between a bettor's expectations and the sportsbook's." For example, in Week 1, the sportsbooks had the Seattle Seahawks as +240 underdogs at home versus the Denver Broncos. +240 money-line odds means that you'd risk $100 to win $240. If you bet $100 on Seattle at +240 and they lose, you lose the $100; if they win, you get your initial $100 back plus $240 in winnings. By setting Seattle's money line at +240, the pre-game odds implied that the Seahawks had a 29.4% chance to win the game ("implied probability"). Positive expected value would exist if you (or some model) believed that Seattle had a higher probability to win than the 29.4% chance implied by those +240 odds. The percentiles referenced below are neither probabilities to win nor confidence levels. They indicate where the expected value of a hypothetical money-line pick falls when compared to the expected value of all potential money-line picks over the last several seasons (i.e., among several thousand possibilities). Teams in Tier 1 ("Selections") are in the 99th percentile or above when compared to all those possibilities. These are rare. Teams in Tier 2 ("Considerations") fall between the 94th and 99th percentile; according to historical back-testing, "considerations" have roughly broken even over time (hence the name). I also include a weekly Top 5 for reference, though betting all of those would have a negative return over time. The model expects to lose more games that it wins, but its selections/considerations are almost always modest to heavy underdogs. Underdogs are underdogs for a reason; the market thinks they are more likely to lose (sometimes a lot more likely) than they are to win. But for those same reasons, money-line underdogs pay more (sometimes a lot more) for victories (e.g., +240) than the amount at risk (e.g., 100). If a bettor's average money-line play is +300, they "only" need to win 25%+ of the time to break even, assuming equally sized plays. The average "consideration" in the period I've studied has carried +343 odds, occurring a little less than twice per week. The average "selection" in the period has carried +488 odds, occurring approximately once every three weeks. (This year's average odds have been much tamer so far.) If a team is listed below, it does not mean the system thinks they are likely to win their game. It means the model thinks they are more likely to win that the current money-line odds imply. And while this column focuses purely on numbers, whether you actually think that's the case is ultimately up to you and your gut! (I always check the Thursday night game, even if I don't publish a given week's article until after Thursday. If a Thursday night team falls in the 94th percentile or above, I will make a brief post and/or email the list.)
This season's top money-line expected-value plays (all 94+s are shown):
1. SEA (+240) vs. DEN - Week 1 - 99.7 percentile - Won 17-16 2. DAL (+270) vs. CIN - Week 2 - 99.5 - Won 20-17 99th percentile threshold separates "selections" from "considerations" 3. HOU (+390) at DEN - Week 2 - 98.2 - Lost 19-6 4. PIT (+380) vs. TB - Week 6 - 97.3 - Won 20-18 5. CAR (+530) vs. TB - Week 7 - 97.2 - Won 21-3 6. WAS (+240) vs. PHI - Week 3 - 96.1 - Lost 24-8 7. SEA (+310) at SFO - Week 2 - 95.9 - Lost 27-7 8. ARZ (+275) vs. PHI - Week 5 - 95.6 - Lost 20-17 9. HOU (+283) vs. IND - Week 1 - 95.6 - Tied 20-20 10. HOU (+205) vs. LAC - Week 4 - 95.5 - Lost 34-24 11. CHI (+265) vs. SFO - Week 1 - 95.4 - Won 19-10 12. DAL (+245) at PHI - Week 6 - 95.1 - Lost 26-17 13. TEN (+115) vs. LVR - Week 3 - 94.7 - Won 24-22 14. PIT (+295) at MIA - Week 7 - 94.5 - Lost 16-10
Searching for Value in Week 8
Last Week - Selections (99th-100th percentile): None - Considerations (94th-99th percentile): CAR (+530) vs. TB: Won 530, PIT (+295) at MIA: Lost 100 All money line; every selection and consideration has been a modest to heavy underdog Season to Date - Selections (99th-100th percentile): 2-0-0 (+515) - Considerations (94th-99th percentile): 4-7-1 (+590) All money line; every selection and consideration has been a modest to heavy underdog The ELDORADO model followed a big Week 6 (Pittsburgh's +380 upset of Tampa Bay) with an even bigger Week 7. On Saturday morning, I notified email subscribers that late-week line movement had pushed two games into Consideration (Tier 2) territory — namely Carolina (+530) versus Tampa Bay (97.2 percentile), and Pittsburgh (+295) at Miami (94.5 percentile). Carolina won big (21-3), while Pittsburgh came up short (16-10). At risk of stating the obvious (or restating what I state every week), going 1-1 on two sizable underdog moneyline plays makes for a very good result. If you combine Selections (Tier 1) and Considerations (Tier 2), the system is now 6-7-1 on the season, or +1105 on 1300 risked, assuming 100 per play. That's an 85% return on dollars risked. (I personally do 1/2 the amount on 94-99s as I do on 99+s, so not quite as good.) And despite my warnings that doing the entire "top five" every week would lose money over time (according to historical back-testing), last week's top five went 4-1, winning +530 (Carolina), +230 (Seattle), +195 (Washington), and +140 (New York Giants), while losing 100 on Pittsburgh. That's a +995 week on 500 risked. Weekly "top fives" are now 13-21-1 season-to-date, or +1075 on 3400 risked, assuming 100 per play. (Some quick asides: You'll note that almost all of that "top five" profit came last week. The aforementioned line movement moved Carolina ahead of Pittsburgh as the top play of Week 7. At +530, they were a 97.2, the fifth highest expected value percentile of the season; they ended up as high as +550 at Caesars, which would have pushed them to a 97.6. You can find a running list of this season's top money-line EVs above.) Week 8 is weird in that every game in the top five is at almost even odds. As weekly followers know well, this system usually sees "expected value" in modest to heavy underdogs. It's also not a very compelling week, as every team in the top five is a low-to-mid 80s play (i.e., not good). None are really even that close to consideration status. So this seems like the type of week to sit back and stand pat. Happy Halloween! This Week's Top 5: Money-line Expected Value 1. NYJ (+120) vs. NEP: 86.8 percentile - The money line implies a 45% chance for New York to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +153 2. LAR (+105) vs. SFO: 86.4 percentile - The money line implies a 49% chance for Los Angeles to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +137 3. TEN (-135) at HOU: 83.0 percentile - The money line implies a 57% chance for Tennessee to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +109 4. NOS (+110) vs. LVR: 82.5 percentile - The money line implies a 48% chance for New Orleans to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +153 5. DEN (+121) at JAX: 82.2 percentile - The money line implies a 45% chance for Denver to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +167
Searching for Value in Week 7
UPDATE (10/22/22): LINE MOVEMENT PUSHES TWO INTO "CONSIDERATION" STATUS (SEE BELOW). Last Week - Selections (99th-100th percentile): None - Considerations (94th-99th percentile): PIT (+380) vs. TB: Won 380, DAL (+245) at PHI: Lost 100 All money line; every selection and consideration has been a modest to heavy underdog Season to Date - Selections (99th-100th percentile): 2-0-0 (+515) - Considerations (94th-99th percentile): 3-6-1 (+160) All money line; every selection and consideration has been a modest to heavy underdog Last week was huge for the model, headlined by Pittsburgh upsetting Tampa Bay as a +380 home underdog. The win swung "considerations" into the black for the year. The Giants also beat the Ravens as a +215 home underdog, which the model saw as having the third-highest money-line expected value of the week. I have not been summarizing the top 5 results here each week, but they are also now in the black. Pittsburgh's win serves as a good reminder that the whole impetus for "Searching for Value" is to remove personal sentiment from the equation. As I mentioned toward the end of last week's post, I really did not like Pittsburgh. But at +380 (more specifically +315 or better), the model said there was 94th percentile expected value or better to be had. So I stayed true to the model and bet it anyway, and I'm glad I did. The win is also a good reminder of the win-loss math that underlies this system. The system's 5 victories this year — 2 as selections (99th percentile or higher), and 3 as considerations (94th to 99th percentile) — have carried average odds of +255. If you win 30% of your +255 wagers, you are in the black. So far this year, the system has won 5 out of 11 of these (with one push), which history suggests is too good to hold up. Week 7 does not currently boast any "selections" or "considerations" (how boring, I know), and its top five money-line expected value line-up is not as far up the percentile chain as last week's. Still, certain teams' odds could trip past the 94th percentile with some upward movement, and at the very least, it's always interesting to see whether/where the model(s), the oddsmakers, and you might disagree. Enjoy Week 7! This Week's Top 5: Money-line Expected Value 1. PIT (+270) at MIA: 92.2 percentile - The money line implies a 27% chance for Pittsburgh to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +290 (worth watching) UPDATE (10/22/22): I was able to get this at +295 on Friday night (FanDuel), good enough to qualify as a "Consideration" (so I took it). The best I am seeing now is +285, just below "Consideration" status. Be on the lookout! 2. CAR (+425) vs. TB: 92.2 percentile - The money line implies a 19% chance for Carolina to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +455 UPDATE (10/22/22): Carolina traded Christian McCaffrey to San Francisco on Thursday, sending their Week 7 odds to +500 on Friday night (where I took them), and as high as +530 today. They are now a "Consideration" (97.2 percentile). 3. WAS (+195) vs. GBP: 90.6 percentile - The money line implies a 34% chance for Washington to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +221 4. SEA (+230) at LAC: 90.2 percentile - The money line implies a 30% chance for Seattle to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +261 5. NYG (+140) at JAX: 84.6 percentile - The money line implies a 42% chance for New York to win - Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +183 (Another interesting occurrence last week came in the way of line movement, as Dallas went from +235 when I posted on Saturday to +245 by Sunday afternoon. If you'd been monitoring the minimum qualification thresholds listed below each top-five entry (as again shown above), you might have noticed that this nudged Dallas into "consideration" territory. I caught it when it happened and backed them (they lost), and I've added them to the running list of "selections" and "considerations" now posted at the top. I also went back and checked whether line movement pushed any other games in or out of qualification status in the early weeks of the season. In Week 1, the Jets settled at +250 vs. the Ravens, down a good bit from an impossible-to-find +275 that had put them in the 94.1st percentile. For fairness and consistency, I have swapped in the Cowboys' Week 6 loss and swapped out the Jets' Week 1 loss from the running list at the top and from the season-to-date "consideration" results. This has a neutral effect on the overall results. Odds movement does not always mean a team should "lose" selection or consideration status in a given week. After all, if you were able to get them at or above the necessary threshold, even if they were only briefly there, that checks the box. The key is whether the "best odds out there" money line that I list in these articles, and which puts them over the threshold, is accessible, or if it's some off-consensus money line that you'd never be able to find. I'll continue to monitor this by posting minimum qualifying odds.)
Searching for Value in Week 6
LAST WEEK - Selections (99th-100th percentile): None - Considerations (94th-99th percentile): ARZ (+215) vs. PHI: Lost 20-17 All money line; every selection and consideration has been a modest to heavy underdog SEASON TO DATE - Selections (99th-100th percentile): 2-0-0 (+515) - Considerations (94th-99th percentile): 2-6-1 (-220) All money line; every selection and consideration has been a modest to heavy underdog Arizona only lost to Philly by a field goal last week, and Cleveland, which earlier in the week met the consideration threshold before line movement knocked them out, lost 30-28. If you're on the email list, I sent a notice regarding that move on Saturday, and I've updated it below. (Going forward, I will note the "odds thresholds" at which a certain team would qualify for or fall out of selection or consideration status.) Your exact odds matter, and they vary by day and sportsbook. Vegas Insider has a helpful cross-book comparison tool (be sure to filter for money line), and I'll usually reference the "most favorable line available" below. But that "best line" is often difficult to find, in part because it might only be temporarily available at some off-consensus book, and in part because you probably only have a couple of apps on your phone. So please pay attention to the "above/below" qualifying thresholds I've now noted below. THIS WEEK'S TOP 5 1. PIT (+380) vs. TB: 97.3 percentile - The money line implies a 21% chance for Pittsburgh to win. Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +315. Becomes a "selection" at +44o or +445. 2. DAL (+235) at PHI: 93.5 percentile - The money line implies a 30% chance for Dallas to win. Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +241. This is close, so the line movement here will be worth watching. 3. NYG (+215) vs. BAL: 92.9 percentile - The money line implies a 32% chance for New York to win. Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +226. This is close, so the line movement here will be worth watching. 4. MIA (+150) vs. MIN: 91.8 percentile - The money line implies a 40% chance for Miami to win. Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +166. This is close, so the line movement here will be worth watching. 5. KC (+130) vs. BUF: 90.2 percentile - The money line implies a 43% chance for Kansas City to win. Do you think they have a meaningfully better chance than that? - Minimum to qualify for "consideration" status is +153. This is close, so the line movement here will be worth watching. I say this all the time, but it bears repeating. The model does not necessarily think these teams are likely to win, and it only "wants" (or historically "expects") to win around "one out of three" or "one out of four" games. But when every pick is a modest/heavy underdog, you can end up in the black even with a losing record. Also remember that offering up the "top five" of the week is more for informational purposes — to let you see how things feel and where your own intuition takes you relative to the models. Betting the full top five would have a negative historical return. Considerations have roughly broken even over time. Selections are fun but rare. My friends often text me begging for "99s." The pressure will be on when another one comes. This is one of those weeks where "gut" and the "models" collide. For example, Pittsburgh feels crazy to me — underpriced, really, like you should be paid more than +380 for accepting that risk. (Although, as a counter to that, 50+ prediction models from across the Internet, summarized weekly by our old friend The Prediction Tracker, give Pittsburgh a 32% chance to win, better than the 21% implied by the current line.) I could see people talking themselves into Dallas, New York, Miami, or Kansas City, but when I floated the initial top five to friends, one hated Dallas ("they can get killed in Philadelphia"), another hated Kansas City ("I'm taking Buffalo to win and cover"), and some won't touch Miami with Skylar Thompson at quarterback. And that's really the point of all this. It inspires some dialogue, and as numbers-based as it intends to be, it ultimately comes down to what you think and feel. Personally, I'll go against my gut and dabble with Pittsburgh, while watching to see if any of the others move to meet the "consideration" thresholds. I might do some of them anyway, even if they don't. Just a little bit. Just to have some fun and skin in the game. Enjoy Week 6!
Searching for Value in Week 5
LAST WEEK - Selections (99th-100th percentile): None - Considerations (94th-99th percentile): HOU (+205) vs. LAC: Lost 34-24 All money line; every selection and consideration has been a modest to heavy underdog SEASON TO DATE - Selections (99th-100th percentile): 2-0-0 (+515) - Considerations (94th-99th percentile): 2-5-1 (-120) All money line; every selection and consideration has been a modest to heavy underdog There are no "selections" in Week 5. To qualify, the expected value of the money-line play must be in the 99th percentile or higher when compared to the expected value of every potential money-line play from the last several seasons. On average, this happens once every 50 games. The two that made the cut this year were Seattle (+245) in Week 1, who upset Denver, and Dallas (+270) in Week 2, who upset Cincinnati. Two teams this week do fall in the "consideration" category — Arizona (+215) hosting Philadelphia, and Cleveland (+135) hosting the Los Angeles Chargers. These are teams with expected values (based on the current odds shown below) in the 94th to 99th percentile when measured in context of the last several seasons. Remember that this percentile range has merely broken even when back-tested historically. I am also presenting this week's "top five;" note that anything below the 94th percentile has had a negative return over time. Below each team, I've indicated their "implied probability" to win based on the current money-line odds. Positive expected value exists when you — whether in your gut or based on some model — believe the team has a meaningfully better chance to win than the current money-line implies. THIS WEEK'S TOP 5 1. ARZ (+215) vs. PHI: 95.6th percentile - The money line implies a 32% chance for Arizona to win. Do you think they have a meaningfully better chance than that? - Falls below the 94th percentile at +195. Minimum to qualify for "consideration" status is +200. (I was able to get it at +215.) 2. CLE (+110) vs. LAC: 89.5th percentile - The money line implies a 48% chance for Cleveland to win. Do you think they have a meaningfully better chance than that? - Falls below the 94th percentile at +130. Minimum to qualify for "consideration" status is +135. (It's come down to +105 / +110.) 3. SEA (+2oo) at NOS: 86.5th percentile - The money line implies a 33% chance for Seattle to win. Do you think they have a meaningfully better chance than that? 4. PIT (+630) at BUF: 84.9th percentile - The money line implies a 14% chance for Pittsburgh to win. Do you think they have a meaningfully better chance than that? 5. DAL (+210) at LAR: 84.8th percentile - The money line implies a 32% chance for Dallas to win. Do you think they have a meaningfully better chance than that? To go from "considerations" to "selections," Arizona would need to move to +285 or +290, and Cleveland would need to move to +200 or +205. Both of these moves would be drastic, and they are unlikely to occur. (Update: Cleveland's best available money line was +135 in the middle of the week. At that level, they met the "consideration" threshold. However, odds that favorable were difficult to find, and by Saturday, they came down to +110 or +105. That line movement knocked them a few points out of "consideration" territory.) Enjoy Week 5!
0 Comments
Leave a Reply. |
|
RSS Feed