2019 NCAA Pool

The Needler

New Member
Dec 7, 2016
1,803
You just made filling out a bracket not fun.
No, the bonuses did that. They operate contrary to their intended purpose, which is why "Erik Honsel is a beast." He's exploiting the mathematics of an inefficient system. Fortunately, this year was an outlier, chalk cruised in the first two rounds, and he got hosed., But people picking like him are likely to win way more times than not for as long as the current bonus system is in place.
 

tims4wins

PN23's replacement
SoSH Member
Jul 15, 2005
37,054
Hingham, MA
This is probably worthy of a separate discussion, but given that (some) people seem to believe there is less variance than ever between the 3 seeds and the 14 seeds, maybe removing the bonuses is a good idea. That would create the incentive to get your bracket as accurate as possible, not to maximize points. And just picking chalk doesn't work because there are invariably upsets, and you need to find them in order to win.
 

The Needler

New Member
Dec 7, 2016
1,803
He's probably correct and he reminds me why I say I am done running this.
I sure hope you don't stop. I enjoy the pool for what it is. I fill out a single bracket, based on my own subjective (and often uninformed) belief of who will actually win each game, and harbor no illusion of ever winning. I bet on the actual games for my gambling fix, and follow my picks in the pool for funsies.

I just do think it's worth pointing out that this format is somewhat beatable, and IMO, the more people who know how, the better. Maybe eventually it'll reach a critical mass so that it becomes not worth it for people to fill out 8 brackets with the 11s and 12s advancing winning in the first round in all, only varying by the 1 or 2 picked to win the tournament. (Or eventually everyone does it, and the guy who had the correct 3-seed wins it all.)
 

Rowan

New Member
Jul 15, 2005
6
He's probably correct and he reminds me why I say I am done running this.
Okay, this is enough to get me--an almost-never-posting lurker--to chime in!

Dan, this is by far my favorite pool and I would be so sad to see it go!

The Needler is making good points about EV of these early round games, but he's overstating the case a bit. Choosing every 12 and 13 to win is a great way to be on top of the standings right now, but can have diminishing returns as the tournament progresses, as you have automatically eliminated viable contenders from reaching the later rounds where the point totals for wins can more than make up for the upset-points in the first round.

The most important factor in winning this pool is picking the correct finals matchup and the correct champion (that's how I won this pool two years ago), which I think most people would agree is as it should be. Getting those early round upsets is a great tie-breaker, and I think it's great that The Needler is pointing out some good math to consider for those who like to take a more measured approach to bracket-filling, but I think it's important to point out that that kind of strategy isn't going to outright win the tournament--it still comes down to picking the right finalists.
 

The Needler

New Member
Dec 7, 2016
1,803
Okay, this is enough to get me--an almost-never-posting lurker--to chime in!

Dan, this is by far my favorite pool and I would be so sad to see it go!

The Needler is making good points about EV of these early round games, but he's overstating the case a bit. Choosing every 12 and 13 to win is a great way to be on top of the standings right now, but can have diminishing returns as the tournament progresses, as you have automatically eliminated viable contenders from reaching the later rounds where the point totals for wins can more than make up for the upset-points in the first round.

The most important factor in winning this pool is picking the correct finals matchup and the correct champion (that's how I won this pool two years ago), which I think most people would agree is as it should be. Getting those early round upsets is a great tie-breaker, and I think it's great that The Needler is pointing out some good math to consider for those who like to take a more measured approach to bracket-filling, but I think it's important to point out that that kind of strategy isn't going to outright win the tournament--it still comes down to picking the right finalists.
I hope I'm not overstating the case. When I say it's beatable, I mean it in the gambling context. That is, it's possible to get a distinct edge, which over a long enough time horizon will generate positive value. In simple and inaccurate terms, if picking the games by hunch, randomly, or some other system would give you a 1 in 300 chance of winning, and this system moves it down to 1 in 30, that is a massive exploitable edge.

You are of course correct that the most points are gained by correctly predicting the winner and finalist. The problem is, there is tremendous bunching among those picks, such that unless you have amassed a large number of points in the early rounds, you're going to lose, even if you correctly pick the winner, unless that pick is a huge upset.

This scoring system values two things over all the others - correctly picking large upsets (which happen almost exclusively in the first two rounds), and correctly picking the winner and finalist. So, while you can't win the tournament in the first two rounds, you can almost certainly lose it there. And the way people do so, in each and every year, is by missing the upsets. When correctly picking a 12-seed to win in the first round is equal to correctly picking EIGHT favorites or an Elite 8 winner, it's pretty obvious what matters is not missing an upset. The reason you won the pool two years ago (Congrats!) is not because you picked #1-seed UNC; I'm sure there were at least a couple dozen others who did so. I'd be willing to bet it's because you had Xavier, Rhode Island, Middle Tennessee State, and USC in the first round.

Also, your diminishing returns point, while superficially persuasive, is not borne out by the statistics. Picking the 11 and 12s to win their first two games only eliminates 3, 4, (each around 10%) 5, and 6 (each around 3%) seeds, who advance to the Final Four extremely rarely. Attempting to correctly pick those rare occurrences is, longterm, an overwhelmingly losing strategy, ie much worse than piicking the 11s and 12s to go through to the Sweet 16.
 
Last edited:

Rowan

New Member
Jul 15, 2005
6
I'll defer to your point about diminishing returns; it was not the correct term for me to use. It's more that there is a non-negligible opportunity cost, and while your EV will be higher if you pick all those upsets, the sample size of 1 tournament each year, plus the (nearly) zero-sum prize structure makes other methods of selection viable contenders. Not sure if I'm making sense.

Another way to think about it: I think most people understand that, if you just went game by game and picked the statistically more likely winner via a reputable source like KenPom or 538, you will have submitted a bracket that is statistically more likely to perform well. Most people don't do that, however, because they find it fun to try and correctly pick games subjectively, and the statistical advantage to doing otherwise often gets washed out in the variance of a 1-tournament sample size each year. It was even mentioned above that if you remove the bonus points, people wouldn't just pick chalk because of course it never pays out as straight chalk. But that doesn't mean it isn't actually statistically a better bet to do that.

What you have done here, which I think is great, is pointed out some of the implications of the particular incentive-structure of this tournament. I think the idea I thought was overstated was that this made the format beatable, which I think some folks would read as meaning you can't win if you don't follow this method, which would certainly take some of the fun out of things!

(for the record, I think I had two of those four upsets, Middle Tennessee and Xavier.)
 

tims4wins

PN23's replacement
SoSH Member
Jul 15, 2005
37,054
Hingham, MA
I hope I'm not overstating the case. When I say it's beatable, I mean it in the gambling context. That is, it's possible to get a distinct edge, which over a long enough time horizon will generate positive value. In simple and inaccurate terms, if picking the games by hunch, randomly, or some other system would give you a 1 in 300 chance of winning, and this system moves it down to 1 in 30, that is a massive exploitable edge.

You are of course correct that the most points are gained by correctly predicting the winner and finalist. The problem is, there is tremendous bunching among those picks, such that unless you have amassed a large number of points in the early rounds, you're going to lose, even if you correctly pick the winner, unless that pick is a huge upset.

This scoring system values two things over all the others - correctly picking large upsets (which happen almost exclusively in the first two rounds), and correctly picking the winner and finalist. So, while you can't win the tournament in the first two rounds, you can almost certainly lose it there. And the way people do so, in each and every year, is by missing the upsets. When correctly picking a 12-seed to win in the first round is equal to correctly picking EIGHT favorites or an Elite 8 winner, it's pretty obvious what matters is not missing an upset. The reason you won the pool two years ago (Congrats!) is not because you picked #1-seed UNC; I'm sure there were at least a couple dozen others who did so. I'd be willing to bet it's because you had Xavier, Rhode Island, Middle Tennessee State, and USC in the first round.

Also, your diminishing returns point, while superficially persuasive, is not borne out by the statistics. Picking the 11 and 12s to win their first two games only eliminates 3, 4, (each around 10%) 5, and 6 (each around 3%) seeds, who advance to the Final Four extremely rarely. Attempting to correctly pick those rare occurrences is, longterm, an overwhelmingly losing strategy, ie much worse than piicking the 11s and 12s to go through to the Sweet 16.
This is a great discussion, and these are great points. Specific to the 11s and 12s. the 6s and 5s you are picking against wouldn't even be favored to win in the second round, so there is very little harm in picking against them aside from losing out on the 1 point - but the 1 point is so inconsequential to the bonus you pick up by nailing an 11 or a 12. And similarly in the second round, you only get 2 points for picking a 4 over a 5/12 or a 3 over a 6/11; but if you nail a 12 over a 4 or an 11 over a 3 that's equal to picking up an extra 4 second round wins. With only 16 second round games, just hitting on 1 of these probably makes it worthwhile. Then you take your regular chalk from there on forward, and yes, you have to get the final four (mostly) right and pick the champion, but like you said there is so much bunching of those picks that the differentiator will be the upset bonuses. But if everyone starts adopting the "pick all 9-12 seeds" strategy, then we might get into a situation where we come down to the tiebreak - which would be a truly ridiculous way to determine this thing, IMO.
 

The Needler

New Member
Dec 7, 2016
1,803
Okay, this is enough to get me--an almost-never-posting lurker--to chime in!
Also, I'm just noticing - you are a perfect example! You did an exemplary job picking winners over the first two rounds. 87% correct, better than the leaders at 85%. Yet because you didn't pick the upsets, you're already 22 points behind, tied for 61st place, and I would guess with very little chance of winning, since there are 35 other players who've picked UVA, including two who are already at least 15 points ahead of you.
 
Last edited:

The Needler

New Member
Dec 7, 2016
1,803
Another way to think about it: I think most people understand that, if you just went game by game and picked the statistically more likely winner via a reputable source like KenPom or 538, you will have submitted a bracket that is statistically more likely to perform well. Most people don't do that, however, because they find it fun to try and correctly pick games subjectively, and the statistical advantage to doing otherwise often gets washed out in the variance of a 1-tournament sample size each year. It was even mentioned above that if you remove the bonus points, people wouldn't just pick chalk because of course it never pays out as straight chalk. But that doesn't mean it isn't actually statistically a better bet to do that.
But with the bonus structure it absolutely isn't a statistically better bet. Take the 8/9 for example - that is statistically as close to a coin flip as you'll ever get. These games are very often pick-ems in Vegas. Yet the 9 gets 2 point if they win, and the 8 gets 1.
 

The Needler

New Member
Dec 7, 2016
1,803
I'll defer to your point about diminishing returns; it was not the correct term for me to use. It's more that there is a non-negligible opportunity cost, and while your EV will be higher if you pick all those upsets, the sample size of 1 tournament each year, plus the (nearly) zero-sum prize structure makes other methods of selection viable contenders. Not sure if I'm making sense.
Well, but people have unlimited entries. I would say an extremely strong strategy would be to complete 8 brackets; in each of them, you piick the 9-12 seeds in the first round, and the 11 and 12s in the second. Otherwise straight chalk, picking each of the 1-seeds in one bracket, and each of the 2-seeds in another.

I'd be willing to do some backtesting if I could get the scoring leaders from past pools, but I'd guess in the great majority of years, this would place at least 1 or 2 brackets very close to the top.

Agian, it's not a guarantee. Some years a 7-seed will win the whole thing. The point is, it's a very significant edge. Think of it like a roulette wheel, but instead of paying you 35-1, they pay you 350-1 when you hit a number. The correct strategy would be to cover every number, but if they don't let you, cover as many as you can for as many spins as you can.
 

Rowan

New Member
Jul 15, 2005
6
But with the bonus structure it absolutely isn't a statistically better bet. Take the 8/9 for example - that is statistically as close to a coin flip as you'll ever get. These games are very often pick-ems in Vegas. Yet the 9 gets 2 point if they win, and the 8 gets 1.
Right--sorry, I was unclear. In this sense I was speaking about NCAA brackets more broadly, not necessarily ones with Murphy's bonus points structure.

My meta-point was just that I think people already have a broad understanding that there are statistical best practices in selecting a bracket that they feel comfortable ignoring in the context of a high-variance, winner-takes-all, contest, maybe because they think they can outsmart the stats, or maybe because they just think its more fun that way.

And to be clear, you are right about every single thing you've posted (including my own chances this year, haha!). I chimed in not to debunk you, but because people seemed discouraged by your points to the point of suggesting altering or disbanding the pool, which I would not like to see.
 

tims4wins

PN23's replacement
SoSH Member
Jul 15, 2005
37,054
Hingham, MA
Right--sorry, I was unclear. In this sense I was speaking about NCAA brackets more broadly, not necessarily ones with Murphy's bonus points structure.

My meta-point was just that I think people already have a broad understanding that there are statistical best practices in selecting a bracket that they feel comfortable ignoring in the context of a high-variance, winner-takes-all, contest, maybe because they think they can outsmart the stats, or maybe because they just think its more fun that way.

And to be clear, you are right about every single thing you've posted (including my own chances this year, haha!). I chimed in not to debunk you, but because people seemed discouraged by your points to the point of suggesting altering or disbanding the pool, which I would not like to see.
But this isn't true for a bonus points set-up. In a non-bonus points pool, if you just picked the statistically most likely to win team in each matchup, you would never, ever win the pool, because there are always upsets - you have to pick upsets in order to differentiate yourself.

Whereas in a bonus points pool, picking upsets becomes almost formulaic because picking a bunch of games correctly in the first round is less important than the bonus points. You absolutely COULD win the tournament with a formulaic approach.
 

The Needler

New Member
Dec 7, 2016
1,803
But this isn't true for a bonus points set-up. In a non-bonus points pool, if you just picked the statistically most likely to win team in each matchup, you would never, ever win the pool, because there are always upsets - you have to pick upsets in order to differentiate yourself.

Whereas in a bonus points pool, picking upsets becomes almost formulaic because picking a bunch of games correctly in the first round is less important than the bonus points. You absolutely COULD win the tournament with a formulaic approach.
Yeah, I'm sure there are some people who understand the significant mathematical built-in advantages available and still pick contrary to those. As I've said, I'm one. But I would guess the vast majority don't. I think if everybody really understood or internalized say that by picking every 12-seed you could never lose more than 4 points to the field in the first round, and could gain as many as 32, or that by picking a #8 seed over a #9 you're essentially saying the 8 is twice as likely to win, you'd see very, very different brackets than the ones actually filled out.
 

tims4wins

PN23's replacement
SoSH Member
Jul 15, 2005
37,054
Hingham, MA
Yeah, I'm sure there are some people who understand the significant mathematical built-in advantages available and still pick contrary to those. As I've said, I'm one. But I would guess the vast majority don't. I think if everybody really understood or internalized say that by picking every 12-seed you could never lose more than 4 points to the field in the first round, and could gain as many as 32, or that by picking a #8 seed over a #9 you're essentially saying the 8 is twice as likely to win, you'd see very, very different brackets than the ones actually filled out.
This post alone just internalized it for me. I'm taking all 9-12 next year, at least in the first round.
 

The Needler

New Member
Dec 7, 2016
1,803
Well, but people have unlimited entries. I would say an extremely strong strategy would be to complete 8 brackets; in each of them, you piick the 9-12 seeds in the first round, and the 11 and 12s in the second. Otherwise straight chalk, picking each of the 1-seeds in one bracket, and each of the 2-seeds in another.

I'd be willing to do some backtesting if I could get the scoring leaders from past pools, but I'd guess in the great majority of years, this would place at least 1 or 2 brackets very close to the top.
For what it's worth, even in this year of nearly-unprecedented 2nd round chalk, each bracket using this system would have 88 points, and of course all Elite 8 teams still alive.

If you chose not to have the 11s and 12s win in the second round and went with the #3 and #4 chalk instead, you'd have 98 points with all 8 brackets tied for 5th, 14 Sweet 16 teams, every Elite 8 team all 8 potential champions still alive.
 
Last edited:

Dan Murfman

Member
SoSH Member
Aug 21, 2001
4,186
Pawcatuck
I guess I know this isn’t the best way to score the pool. But It’s been the format for so long it would tough to change. It will something I will I think about for next year.

I will say if it wasn’t for finding Venmo I very much doubt I would still be running this. The year PayPal froze my account I really was ready to say that’s it. Another issue coming up is the software I use and they also host the pool is saying they’re not sure how much longer they will be offering it. I’m sure there are others but I’m not looking forward to that.
 

The Needler

New Member
Dec 7, 2016
1,803
I guess I know this isn’t the best way to score the pool. But It’s been the format for so long it would tough to change. It will something I will I think about for next year.

I will say if it wasn’t for finding Venmo I very much doubt I would still be running this. The year PayPal froze my account I really was ready to say that’s it. Another issue coming up is the software I use and they also host the pool is saying they’re not sure how much longer they will be offering it. I’m sure there are others but I’m not looking forward to that.
Yeah, I’m sure it’s a hassle, and it is appreciated. You should add a small administration fee. 50 cents or a dollar for each entry shouldn’t be objectionable to people.
 

DJnVa

Dorito Dawg
SoSH Member
Dec 16, 2010
53,840
All I'd say, is if you do decide not to continue let us know. Perhaps someone would be willing to take over, and could use your email list, etc. and keep it going.

That said, I hope you keep rolling with it.
 

Dan Murfman

Member
SoSH Member
Aug 21, 2001
4,186
Pawcatuck
I’m sure I will. Im just a little down that this is the first year that lupojohn didn’t email me and ask if he could get in.
 

Dan Murfman

Member
SoSH Member
Aug 21, 2001
4,186
Pawcatuck
I don't understand how "worst finish points" is nearly always higher than "current points." If somebody picks every game incorrectly going forward, those numbers should be the same, no?
I received an answer. Supposedly in the one that says what's the worst finish place it gives what's the worst position you'll wind up in. And worst finish points tells you how many points you will have if you finish in that place.
 

The Needler

New Member
Dec 7, 2016
1,803
I received an answer. Supposedly in the one that says what's the worst finish place it gives what's the worst position you'll wind up in. And worst finish points tells you how many points you will have if you finish in that place.
Yeah, that still doesn't make any sense to me, though I'm not sure it matters.

Never mind, I get it. I could finish in a better position despite scoring fewer total points.
 
Last edited:

The Needler

New Member
Dec 7, 2016
1,803
For what it's worth, even in this year of nearly-unprecedented 2nd round chalk, each bracket using this system would have 88 points, and of course all Elite 8 teams still alive.

If you chose not to have the 11s and 12s win in the second round and went with the #3 and #4 chalk instead, you'd have 98 points with all 8 brackets tied for 5th, 14 Sweet 16 teams, every Elite 8 team all 8 potential champions still alive.
By my calculations under this system, you'd now have all 8 brackets with 108 (picking the 11 and 12s two win two games), or 118 (picking them to lose in the second round) points. In 5 of your 8 brackets you'd still have the champion alive, and a potential to score as many as 218 points total (if MSU beats UVA or Kentucky beats Duke or Gonzaga in the final).
 

Dan Murfman

Member
SoSH Member
Aug 21, 2001
4,186
Pawcatuck
I just got notified that my pool software company is still in business so I bought the update so the pool goes on. I must admit as the years go on I really not in love with the scoring though I will say my local people say they enjoy the underdog scoring. So I am curious what people think is the best way to score a March Madness pool. But I guess after 25 years it might be hard to change it.
 

tims4wins

PN23's replacement
SoSH Member
Jul 15, 2005
37,054
Hingham, MA
I just got notified that my pool software company is still in business so I bought the update so the pool goes on. I must admit as the years go on I really not in love with the scoring though I will say my local people say they enjoy the underdog scoring. So I am curious what people think is the best way to score a March Madness pool. But I guess after 25 years it might be hard to change it.
Seems like it has become too formulaic in terms of the upset points. I think it would at least be productive to discuss different options.

I’ll throw out an option:

First round games worth 1, second round games worth 2, third round games worth 3... champion worth 6. That places a huge premium on getting as many picks correct as possible.

For upsets maybe just give a 1 point bonus regardless of seed. So again you have to be really strategic.

In general I am in favor of whatever rewards picking the most correct games and not getting like 60% right but nailing a bunch of huge upsets and the national champ.
 

DJnVa

Dorito Dawg
SoSH Member
Dec 16, 2010
53,840
I was just reading article about this, by a site that runs a lot of pools.

50% of pools run the 1-2-4-8-16-32 system. It's considered easier for those that aren't too knowledgeable about the picks.

The second most popular is only used 5% of the time--it's 1-2-3-4-5-6, like said above. It makes skill important in the early rounds and picking the champ right is worth only 5% of the total points, which means you can easily have winner that doesn't have the champ picked correctly. Tough to make up ground late if you didn't do well early.

There's the 1-1-1-1-1-1 system--the first 2 days are worth more than rest of games. It's the one the most rewards skill in picking.

There's an interesting on called the Fibonacci system: 2-3-5-8-13-21. This one emphasizes the first 2 days, but still gives players chance to catch up later. It rewards those that pick well late, but you have to score well early to be in it.

Then they profiled a random one: 2-3-5-7-10-50. Most points available in first 2 days, but you gotta nail the champ.
 

HomeRunBaker

bet squelcher
SoSH Member
Jan 15, 2004
30,096
I just got notified that my pool software company is still in business so I bought the update so the pool goes on. I must admit as the years go on I really not in love with the scoring though I will say my local people say they enjoy the underdog scoring. So I am curious what people think is the best way to score a March Madness pool. But I guess after 25 years it might be hard to change it.
The only change I'd recommend is capping the number of points an underdog can earn. Ex: Instead of a 13 beating a 4 earning 9 points it is capped at 5 so an 11 over a 6 would count the same as a 14 over a 3. This allows for more actual handicapping than simply playing the analytical game by picking all the 12, 13 and 14's in the first two rounds.
 

Dan Murfman

Member
SoSH Member
Aug 21, 2001
4,186
Pawcatuck
I like HRB suggestion but not sure if program can do that. And I'm thinking let's just try one year without and see how it goes. And I agree with Tim that it should be who picks the most games. But I do also think picking the champion should be rewarded too. I kind of like the 2-3-5-7-10-50 example.
 

tims4wins

PN23's replacement
SoSH Member
Jul 15, 2005
37,054
Hingham, MA
I wouldn’t mind losing upset bonuses altogether. We’re past the age of 12s and 13s being true underdogs. Removing the upset bonus wouldn’t result in all chalk picks because there is no chalk.

I do understand rewarding picking the winner. But I think 50 is way too high.

Under the 1-2-3-4-5-6 method, there are 120 total points. Picking the winner earns you 21 or ~17%. Picking the Final Four, the title game, and champ gets you 56 out of 120.

I don’t know what the right formula is but it does seem like there is a happy medium somewhere. 1-2-4-6-8-10?

Under that scenario there are 146 total points. Picking the winner is worth 31. If you nail the F4, title game, and champ you earn 78 out of the 146.

1-2-4-6-9-12: 150 total. Champ worth 34. F4, title, and champ worth 82 of 150.
 

The Needler

New Member
Dec 7, 2016
1,803
I don’t think there’s anything too wrong with the point system other than the bonuses for upsets being disproportionate and so incentivizing just picking the underdogs in the first round. This can be remedied by increasing the starting points awarded for a win. Instead of 1 point for a first round win, you make it 5 (somewhere around 7 is probably optimal, but 5 would probably be good enough to discourage the all-upset first round strategy).Then keep doubling with each round as before, and keep the bonus points as before. So now if you correctly pick an 11 beating a 6, you get 10 points, and if you correctly pick the 6, you get 5 points, which is a much better approximation of the chances of winning (2:1).
 

tims4wins

PN23's replacement
SoSH Member
Jul 15, 2005
37,054
Hingham, MA
I don’t think there’s anything too wrong with the point system other than the bonuses for upsets being disproportionate and so incentivizing just picking the underdogs in the first round. This can be remedied by increasing the starting points awarded for a win. Instead of 1 point for a first round win, you make it 5 (somewhere around 7 is probably optimal, but 5 would probably be good enough to discourage the all-upset first round strategy).Then keep doubling with each round as before, and keep the bonus points as before. So now if you correctly pick an 11 beating a 6, you get 10 points, and if you correctly pick the 6, you get 5 points, which is a much better approximation of the chances of winning (2:1).
This could work, for sure.
 

Dan Murfman

Member
SoSH Member
Aug 21, 2001
4,186
Pawcatuck
Okay I was going to go with keeping the underdog points and going 6 12 24 48 96 192 but the program only allows up to 99 so I can’t do 192.