If you can get a +10% advantage to win in anything you do and you don't do it based upon your gut alone.....well your gut may be fine, it's the brain that is the issue IMO.
The math behind going for 2 when down by 14 late in game and scoring a TD.
- Thread starter Grant Green
- Start date
You’ve eliminated the possibility of still winning the game when you miss your two pointer.
I think we are actually agreeing and maybe I worded my response poorly.
Hypothetical - If you are an outlier team where your 2 pt conversion rate is really poor (lets say 38%), but you think your chances of winning in OT are really good (lets say 58%), then in theory you would be better off playing for OT.
And I'd argue that if those are your assumptions, then you're probably doing something wrong in your calculations. It goes back to what you said...that 10% to 15% is such a big swing that things like momentum, struggling so far this season in short yardage, etc are just not strong enough to overcome an advantage that big.
I agree, which is why I made it a theoretical. In reality I don't think there are teams with that large of a discrepancy between 2pt success and OT success.
ahh, but how would you know there is a 10% advantage? Its not like baseball where you have a large sample size of pitches, plate appeaances, etc. You only got 12 games and at least half of those are against patsies.
We do agree, I can see how the two success rates would move together but there could be a difference even based on style of play. Factoring in differences there plus your defense and opponent and kicking, and the momentum/morale piece which might be overplayed but shouldn't be ignored, and the picking up tendencies or wearing down a defense, there's a lot of ingredients in the cake, enough that it's not simply a 62.5% > 50% exercise. But I love the conversation and agree with the overall logic and ultimately won't fault any coaches for attempting 2pt conversions in those scenarios.
Well, I think we are 95% in agreement, but going in circles on the other 5%.
I'll just leave it at this - while all of these factors are worthy of consideration, they just can't overcome an enormous 12.5% edge. In fact, many of them could cancel out (ex. you could in theory lose momentum by missing the 2pt, but gain it my making it). Even if the edge gets reduced to 5%, I'll take it every time.
We 100% agree that this is a fun topic!
I love discussing this stuff, so thanks for starting it. What's really interesting is that you always hear all of these justifications and reasons looking to not go for 2. You never see a coach go for 2, or go for a 4th down, in a spot that makes everyone say "well, it went against the analytics, but it was a good call". An example was Franklin going for it in the first half against Michigan. I didn't like the decision, but he seemed to think he had a reason to go for it there...I guess if it were successful, you'd have more people on board, but I'm never one to get involved in results-based thinking. But really, his decision to go for 2 didn't go THAT strongly against the analytics if he thought it was really going to be a low scoring, low possession game. It seems that everyone is mostly in favor or sticking with the practices that have been commonly accepted over the past generation, and that deviating from them opens the coach up for criticism even if there is a good reason in doing so. And often, "extending the game" (or as we touched upon earlier in this thread, "extending the series") seems to play way too much into people's thinking, if the end goal is actually "winning the game".
And since I mentioned it, I went back and grabbed the Norv Turner blackjack story...it's the start of this column. Still cracks me up reading it again...
The Sports Guy: Attack of the Super Blog
Bill Simmons is onsite at Super Bowl XXXVIII -- and he's filing round-the-clock reports from Houston.
www.espn.com
Yep to all of that. Confirmation bias and sports dogma are difficult for people to overcome.
I would like to know the math about going for 2 when ahead by 14 late in the game! Ya never know, it may happen some time!
Ugh, Norm.
I haven't played a lot of blackjack in my life, so curious to get your thoughts on this. Hopper complains about Norm not hitting and then getting his crappy card. I've seen people complain about this before, when a person makes the wrong move and changes the card they get. This always seemed like faulty logic to me. Isn't there just as much chance that they would have taken a card that would benefit you?
Yep, but people still get fired up about it…it’s like the whole 2 point stuff, in that when it helps you, you never think about it or remember. But hurts you, and it’s a big deal.
Ok… up by 14, 3 minutes to go and we score. Do I open a Stella or do I do a Captain Morgan and coke?? There has to be analytics on this!!
Stella now and then Morgan and coke when Franklin scores again because he wants to cover the spread.
That's where you need to take the time into account...night game, and the Captain and Coke is the way to go. Noon game, and that 3pm Captain Morgan never leads anyplace good. 3:30 game could go either way, so trust your gut based on the momentum of your day.
You had fun composing that one, didn’t you?
Or you just ignore the analytics and go straight to the thinking mans drink. It's a "gut" call.
I'm not sure about that.
Beer before liquor...
It was more fun than the relatively boring FA Cup games that I was watching this afternoon.
Agree with this, but if you go into a decision with analytic modeling saying you'll have a 50% success and fail 7 times in a row, did you get really unlucky or was there something wrong/incomplete with your model?
From my experience, models can be very sensitive to additional dynamics being introduced especially when it's a newer model or newer data. I'd love to dig into the stuff Penn State uses and I'll definitely check out the content recommended in this thread.
AL BUNDY: “I only watch sports where something happens!”
At 7, you’ve still got a good chance of it just being variance at play… just not enough of a sample size to gain anything meaningful from it.
After 2, the chances of your model being materially inaccurate grow exponentially. Possible that 50% is still accurate but very unlikely. Chance of failing 7 straight times on a 50% success rate is less than 1%.
One year in Vegas, my buddy got the idea of finding the roulette tables that hit a bunch of the same color in a row (probably 5+) and playing it. Obviously figuring that the odds had to be super low that it would be a 6th, or 7th of the same color he bets against it. If it missed, he doubles his bet and keeps doubling until he hits (not realizing that this is the Martingdale system). I tried to explain to him that his odds were still 50/50 regardless of how many had hit in a row, but he wasn't convinced. Of course some other guys bought into the idea. They lost a lot of money that weekend...never did it again.
Your boys have it all wrong…my friend that boasts of winning often in AC is sure that the key is to find a hot table, and know when it starts going cold to walk away and find another table.
Yeah, that's not how probabilities work but your buddy probably represents the majority of the country: "It's due!!"
Modeling for roulette is about as simple as it gets, theortetically zero variables so you can be very confident in the 47.4% black/red odds regardless of what table you're playing. Coin flips are 50%, easy. With football we could come up with dozens of variables that should be considered which is why I wouldn't (necessarily) use the national average success rate for my team's probability on 2pt. If were flipping coins, yes it's possible to fail 7 times and the 8th attempt would still be 50%. But in real life football with a specific offense/defense, each successive failure indicates my team/skiddish qb/dropsy Wrs/etc might just not be very good at it and we should kick the xp to maximize our chances at a win. Hypothetically, of course.
Back to the original post. I think many are not addressing, when is ‘ late in the game’ ? 2 min to go? End of the third quarter? 9 min? 7min? That imo this is the key, deciding when ‘ late in the game is’?
I’d be pointing towards anything from about the 10-12 minute mark on, but game flow does have an impact. If it’s a shootout where neither team can stop each other, you wait longer than in a 14-0 slugfest where you’re expecting scoring to be at a premium.
Coin flips are not 50-50... I have a good story about this.
Coin flip at the middle of the muddy field. Ref did not catch the coin, but let it hit the ground. Didn't that coin land vertically on its edge. maybe 1 in a million.
Greg Olsen thoughtfully tore apart Mike Wilbon’s rant on NFL announcers using more analytics
Greg Olsen blasted Mike Wilbon’s analytics rant in the most polite way.
ftw.usatoday.com
Here’s a good article about how some mainstream sports commentators don’t quite understand the math behind all of this. I don’t deny any of the math proven by the gentlemen in this thread, but I will say that as a long time coach, the lower the level that you go the more that other variables like momentum and home field play into it. When you get to the NFL, these are the best players in the world and you basically know what you have and the outlier outcomes that some are suggesting are not something that people should even consider.
The only variables with this going for two scenario is what you believe the time left is for you to get that third possession and potentially an injury to one of your necessary players.
Last edited:
Old post, but I'll respond ... the thing there is that Norv was making the correct move, and the "Hopper" fellow is a mathematical idiot. You never hit a 16 when the dealer is showing 3. It's not even a close decision. If you were at a table filled with non-mouth-breathers and a guy like Hopper pulled up and started hitting 16 on a dealer 3, he'd get run off in short order.
The other night Ravens v Chiefs. Likely scores on last minute TD ( later over turned so none of this matter) the Good Harbaugh immediately signals go for 2 to win the game. I thought you just battled you *** of for 60 min coming from behind, why win or lose the game on 1 play??? Too much effort put in to lose. So kick the xpt head to TOGreg Olsen thoughtfully tore apart Mike Wilbon’s rant on NFL announcers using more analytics
Greg Olsen blasted Mike Wilbon’s analytics rant in the most polite way.
Here’s a good article about how some mainstream sports commentators don’t quite understand the math behind all of this. I don’t deny any of the math proven by the gentlemen in this thread, but I will say that as a long time coach, the lower the level that you go the more that other variables like momentum and home field play into it. When you get to the NFL, these are the best players in the world and you basically know what you have and the outlier outcomes that some are suggesting are not something that people should even consider.
The only variables with this going for two scenario is what you believe the time left is for you to get that third possession and potentially an injury to one of your necessary players.
