Quote: Originally Posted By smackawits on 4/20/2010

Quote: Originally Posted By schroedess26 on 4/19/2010
Realistically though just because it did not happen in real life does not mean it couldn't happen. Its just that much more unlikely. A 16 has not beat a 1 but if we played another 300 hundred years I would bet that it would.

Michigan State 69, Alabama State 35 (South regional
first round, 2001) 29-25 halftime lead
Michigan St. went on a 26-0 tear and outscored the Hornets 40-10 after the intermission. Alabama State shot just 13 percent from the field (4-for-30) in the last 20 minutes.

November 3rd, 1978 Duke lost to instate rival Davidson 49-33. Amazingly, Davidson came back from a 31-4 halftime deficit. This is the fewest number of points ever scored in a half of NCAA men's basketball.

December 14, 2009
Bethune-Cookman vs Carver Bible 35-28 at half and 99-55 Final
Bethune outscored Carver 64-27 in the second half

December 30, 2001
LSU vs Nichols State
22-25 at half with Nichols State Winning
LSU then outscores Nichols state 49-10 in the second half and wins 71-35

December 30, 2000
New Mexico vs Texas-El Paso
43-42 Texas-El Paso Winning at half
121-86 Texas-El Paso Wins the game outscoring New Mexico 78-44 in the 2nd half

March 16, 2007
Florida vs Jackson State
Florida winning 41-35 at half
Florida wins 112-68 scoring 71-34 in the second half

There is quite a few games and thats just with a quick look at the last 10+ years.

I notice that colon-els hasn't responded to this one yet.....probably won't....it's not what he wants to hea

Technically, none of those games meet the criterion that I asked for (tied at halftime, 40 point swing in 2nd half), just wanted to point that out to begin with. The Duke-Davidson game is close and noteworthy (the only game where there was a 40 point swing) , however you can make a strong "coasting" argument there. With that said, I have a lot of respect for schroedess for going out and seeking the information when he wasn't even the one that was called out here. I called out aintheb here, and he's so lazy that the only thing he did was type in "Largest NCAAB comeback" at Google and returned a result that was rather irrelevant to the conversation, as dalter mentioned/agreed with.

Again, if the 50-9 second half result in HD is justifiable, then do it...if you're saying that this should happen within the realm of the sim, then I would love to see your argument or anyone elses for that matter.

Ohio Valley went +41 in a 19:41 span (19:45 of the 2nd to 0:04 of the second) on the ROAD and only had a 16 point adjusted OTR rating (minus WE, ST, and DU.....Ohio Valley's is 369.... Barry's is 353) advantage on their opponent. No walk-ons for either team. Yes Barry switched from 0 to -2 on their FCP...but does that justify a 41 point swing, especially when they were tied at halftime? Its not like they switched to -5.

Well I assume it was partly bad luck here with Barry just happening to miss shots more. Also, coasting would seem to be the biggest reason that these blowouts don't get bigger usually.

I think the biggest problem is that 60-70 points in a half is not scored that much in real life. In HD everyone scores much faster with uptempo and normal pace as opposed to real life. You need to score 60-70 cause even bad teams usually will score 20 in a half in real life.

In the end the starters would have come out of the game with 2-5 mins making them score most likely 5-10 points less. It ends up a +31 point second half and even though its still a lot its something that has been see a couple times a year. Not when teams are tied but within a couple points at half time.

ok colonels, here is your argument.

lets simplify the sim to a somewhat comparable model to examine the effect of randomness on the outcomes.

in the sim, not that many shots are taken by a team in a half. there are also a whole bunch of other coin flips, but many have less weight than each shot. also, there are 2 and 3 pointers. but to simplify, lets say there are 80 1 point shots being taken, and no other randoms involved. in reality, 2 and 3 pointers are more volatile, but things like, which defensive player tries to get the rebound, are less volatile. so this seems like a reasonable middle ground.

so, we have 80 equal weight coin flips for each team. team 1, the superior team, is shooting about 50% fgs against a weak schedule. his opponent is somewhat weaker than his schedule, but he was on the road, so lets stick with 50% for his probability of making each point. team 2, the inferior sim team, is shooting 42% against a significantly weaker schedule than his opponent, who crushed him. so lets go with 35%.

anyway, the expectation of team 1 is to make 40 shots, for 40 points. the probability of this event is 8.8%. The probability of the event in the sim, of team 1 making 50 shots, is .7%. that's a little under 10 times as likely as the expected outcome. so really, nothing surprising there.

the expectation of team 2 is to make 28 shots. the probability of this event is 9.3%. in the sim, team 2 only makes 9 points, which has a probability of 9.5 times 10^(-7). this is approx 100 thousand times less likely than the expected outcome.

with about 10K games simmed per day, and 20K halves, i don't think seeing an outcome that is 1 in 100K is particularly surprising. with 20K halves, i would expect a larger outlier than that on a daily basis.

as we have discussed before, i think it is extremely difficult to hang your hat on any single example and proclaim the sim is broken. if we took a sampling of say a hundred or thousand games a day for 10 days, and looked at those results, then an argument might be able to be made that the occurrence of unlikely events was not in line with what you would expect from simple randomization of every event.

i personally would not be surprised to see there were too make unlikely events. i am kind of on the fence, wouldn't be surprised either way, maybe even leaning towards expecting "too many". and my best guess would be that it is a result of a random factor that applies to more than a single coin flip, which would certainly skew the results from when we would expect if it was significant. for example, if there was a random factor on the weight of HCA which was applied to a large number of the coin flips in a simulation, it would throw off the normal bell curve we would expect considerably (IMO at least).

also, at the 12 minute mark when barry switched to uptempo, all bets are off. the other team's guards were much faster, playing a + defense, no telling if barry will give up countless turnovers and take horrible shots (really anything on the per is going to be bad, not only was it bad to start against better defense playing a +4 man, but now they are rushing the shots) or not, but its very likely that will happen. so the final 30-7 run is not very meaningful or surprising, IMO.