Edit:
a complete list of all the dynasty lists, longest streak lists, and others can be found in the dynasty list index
dynasty list index and more


For those who don't know, I want to make a cross-world dyansty list, both current and all time. I have the data all ready to go, as you can see in my first crack at an all-time streaks list by team (for example, most consecutive NT appearances of all time for a program, by division) here.

Anyway, I would like to get some feedback on how to do the dynasty ranking. This isn't meant to be a definitive list of the greatest programs or anything like that, just for fun. But, I'd like to get as much community input as possible so I can use a system that most people feel is a good one.

I'll start with the concepts I am thinking of using now, and get into specifics later. Most generally, I am thinking of doing a dynasty ranking over a period of years, with each year weighted equally. What do you guys think for period of time? I am thinking 10, and maybe also 20 for the all-time dynasty list, and 10, and maybe also 5 for the current list. Also, how do people feel about weighting each year equally? I feel that is best for the all time list, but am torn on the current dynasty list.

In terms of actually ranking the teams, I was going to use a sum of the seasons approach. For each season, there are a few different concepts I am considering:

1) Base team rank - this would be a combination of just rpi and post season success. I will post the formula I came up with for first draft, but I think this is a very important one to get buy-in on. It would be the starting point for everything, but is simple enough that hopefully we can fairly easily hash out a good way of weighting rpi and post season success.

outside of base rank, there are still all these factors - records (overall, home, away, neutral, conf), as well as conf champion status, and conf tourney success. SOS is not available on the history page, so there is no way for me to get that data retroactively. i feel like given rpi and post season success, the number of wins in a season is fairly irrelevant. at least, in many cases, it doesn't seem clear to me if more wins is better or worse (knowing there is a corresponding worse or better SOS). what do you guys think of that? i was thinking maybe a small bonus for 20 and 30 wins or something but really i feel that is pretty arbitrary too.

well, that leaves conf champ, and the CT. because there are so many conf champs, and it is so much based on the strength of your division, i figured, screw conf champs. but the CT is a pretty big deal, a NT bid or a big effect on seeding is on the line for everybody. i love the CT and they can be very competitive so I wanted to include them. but, the CT champ in the top conference getting the same value as the CT champ in the bottom conf is obviously flawed. so that brings me to

2) Base conf rank. I think a way to rank conferences in a primative, single season way is necessary to use CT success as a factor in the dynasty rankings. Also, one of the other dynasty lists I'd like to generate is a conference dynasty list, which would certainly be more complex than this base conf rank - but I just wanted to mention that because there is a pretty cool list IMO (conf dynasty list) that can be generated as soon as I have a good way of ranking conferences. Anyway, back to base conf rank... I basically was thinking of summing the 12 base ranks of the teams in each conf. And then like, some how scaling conf tourney success. How to do that is not clear to me, I'd love some ideas. Like, do you set the top conf worth a base of 2 (or 3 or 4) times the lowest, and put everybody in between linearly? That doesn't seem best, because in different seasons, the gap between first and last changes. Maybe when we have the base team rank, I can pull the mean value and the variation, and we will then have the data to make a decision on conf rank?

3) Team rank - this would be a combination of base rank, and conf tourney success. how do people feel conf tourney success should fit in the equation? like, for starters, is a CT champ in the best conference compared to an early exit in a moderate conference as significant as a 1st round NT loss vs 2nd round NT loss? i personally think winning the hardest CT is pretty tough and would put say a hardest conf CT title and 2nd round exit probably in the ball park of on par with a no show CT and a sweet sixteen. i could definitely see it being better too.

4) World rank - this is the last piece, and a critical one. a string of 10 good seasons in the most competitive d3 of all time has to be worth a hell of a lot more than the same performance in a dead d3 conf in one of the new worlds, right? Well, I really think an objective standard can be found. For clarity, when I say world rank, I really mean division rank, like d3 Tark rank. But with conf rank, I thought division rank would be confusing (could be the div of a conf), so I am going to call it world rank - but its really division of a world rank :)


OK, well those are my basic ideas on where to go with the dynasty ranking. Now, here is my first crack at actual numbers for base team rank.
There are 2 components, rpi and post season success. For rpi, I assigned point values for different rpi points, and plan to scale rpi values in between the defined points linearly. This is not perfect but I think it works out pretty reasonable :) Here are the point values I assigned, and the rpis associated with them:
10 - 1
9 - 2
8 - 5
7 - 10
6 - 17
5 - 26
4 - 37
3 - 50
2 - 65
1 - 100
0 - 150

Basically, what I am looking for on this is how do you think the curve fits? To test the curve, I tried to think of some examples. Like, if you have 2 seasons, and get a #1 rpi and a #25 rpi, what rpi value would you have to repeat to tie that? Well, #1 rpi and #25 rpi together in my set of numbers come out for be roughly 15 points. That corresponds to getting an rpi of 7 one season and 8 the next. I thought those were fairly equal, so in my mind, this point system passes that test. If you guys could let me know if you think the system breaks down, and some data points where, that would be excellent! Suggestions for minor changes would be great to - it doesn't have to be totally different, I want to get this "optimized" if you will to a pretty fine grain. One final detail on rpi, I do let it go negative. So the #200 rpi team would get a value of -1, because that is the linear interpolation off of rpi 100 = 1 and rpi 150 =0. So #300 rpi team would get a value of -3. I kind of did it this way because I felt like crappy seasons are crappy seasons and didn't really care much how crappy they were. So the weight difference of a pretty crappy vs really crappy season is very small compared to a pretty good season and a really good season.

For the second piece, post season success, I assigned values for every possible outcome. Forget the PIT for now. For the NT, I felt like if the first round was worth X, then the value of the sweet 16 should be roughly 2X. And it should double every 2 wins. So the final four would be worth twice sweet 16, and the championship twice the final four, etc... As a result, for example, a s16 and final4 are worth more than 2 elite 8s, but not dramatically (which I think is right). How do you guys feel about that basic concept, the doubling of the value every 2 wins?
Well, I applied this to some numbers. Here they are:
NT 1 loss - 3
NT 2 loss - 4.2
S16 loss - 6
E8 loss - 8.5
F4 loss - 12
Title loss - 16.9
Title win - 24

So, there are really 3 questions for base rank in my mind. Is the score for rpi internally consistent (meaning just for rpis)? Is the score for NT success internally consistent? And are the weightings of the two fair? I've already explained my reasoning for the first two, so now onto the last :)

To try to balance rpi and NT success, I kind of looking, what rpi would I expect from a team who has a performance of _blank_. So, for example, NT 1 loss is worth 3, so is rpi 50. That seemed pretty reasonably to me. The NT 1 is probably a little better, but its close. I felt having rpi and NT success being equally weighted was a fair way of valuing the regular season and the post season, especially at the lower tiers of NT teams when there is SO much volatility and luck in your post season outcome. I felt if anything, at the lower end, regular season should be weighted heavier, and at the higher end, post season success. For another example, a sweet 16 is worth 6, so is rpi 17. The average rpi of the team losing int he sweet 16 is 12, which is worth a little under 7 in the rpi system, so I thought that was again, pretty reasonable, being in the earlier stages of the NT. Then things kind of blow up. The #1 rpi is worth 10, which is better than an elite 8, and worse than a final 4. I thought that made perfect sense, because honestly, once you make the final 4 the difference in what your rpi is starts to mean crap. And I'd rather have a final 4 loss than have the #1 rpi any day of the week, even if a final 4 loss puts me at 3rd or 4th in the country.

My main concern is after the final 4. A championship is worth 24, which sort of dwarfs rpi differences. But maybe it should? How do people feel about that one? I honestly think the metric of a championship = 2 final fours is a very reasonable one. But it seems like it might be a little too high with respect to rpi difference. At other times I feel its perfectly reasonable. So yeah... what do you guys think?


I don't have any figures yet for base conference rank, or team rank. I figured those would make more sense after we could agree on a format, and a base team rank :)

For world rank, I have come up with some numbers, but I will spare you guys right now because its not very polished. I just wanted to elaborate a bit on the format I am thinking of using, to see if you guys agree there is conceptual merit to a world ranking metric.

My idea is to assign a value to every school in a division, and to use the division sums somehow to affect the weighting of that season's success. There seem to be 2 steps to me - first, coming up with a ranking system that is internally consistent, and second, figuring out how to translate that into a weight on a season's success. Let's focus on the first part for now - I think that is definitely the harder part anyway.

I am basically thinking something like this: the same human coach at a school for 10 straight seasons is worth 10 points. a simAI at a school for 10 straight seasons is worth 0 points. for the rest of the scenarios, i would use some value in between. there are literally a shitload of possibilities, like 5 coaches coach for 2 seasons each. but i think a heuristic could do something reasonable with all of them. i also feel like 10 straight seasons is where there is basically 0 return on additional seasons, in terms of that school's dynasty potential. that is why I made 10 seasons the max. but i feel the large majority of the impact is made after 5 seasons. so a simAI for 5 seasons with a human for the 5 before that would probably be worth like, half a point. and a human for 5 seasons following a simAI for 5 might be worth like, 8 or 9 points. a human for 5 following another human for 5 would probably be worth 9.5 or more. something like that. any thoughts?

I've used a few different rating systems.

Awarding points for postseason I'm pretty close to the list you gave; I award 2 points for making the NT and then 3 points for each win. So the points are 2-5-8-11-14-17-20. Maybe not heavy enough towards the champion. PI is 1 point for making the PIT and 1 point per win = 1-2-3-4-5-6.

For season prestige I average the ending RPI with a rating associated with postseason performance.

Lose in first round of NT between (33rd-64th team) = 48.5 rating
Lose in Second round (17-32) = 24.5 rating
Lose in Sweet Sixteen (9-16) = 12.5 rating
Lose in Elite Eight (5-8) = 6.5 rating
Lose in Final Four (3-4) = 3.5 rating
Lose in Final = 2 rating
Championship = 1 rating

iguana, you talked about post season points, and season prestige. are those 2 different systems? or are those two things you used together? if you used them together, can you explain that bit

countdown to post by Colonel Mustard explaining his systemm.......

I think what you describe is cool - and my personal view is that I would not sweat the details - would be interesting although that is a ton of work!

I think recent seasons should be given more weight in dynasty rankings. Let's say you used the last 20 seasons, and I'm going to use an extreme example; Team A won 5 national championships in seasons 1-5 but only made the postseason twice in seasons 6-20, while Team B only made the postseason twice in seasons 1-15, but won 5 straight national titles in seasons 16-20. Team B would be looked at as the more prestigous program. I don't think it should be hugely weighted, but maybe something like the last 5 seasons being worth 20% more than the previous 15?

Quote: Originally posted by kmasonbx on 2/16/2010I think recent seasons should be given more weight in dynasty rankings. Let's say you used the last 20 seasons, and I'm going to use an extreme example; Team A won 5 national championships in seasons 1-5 but only made the postseason twice in seasons 6-20, while Team B only made the postseason twice in seasons 1-15, but won 5 straight national titles in seasons 16-20. Team B would be looked at as the more prestigous program. I don't think it should be hugely weighted, but maybe something like the last 5 seasons being worth 20% more than the previous 15?

how do you feel about this in the context of an all time dynasty ranking, vs a current dynasty ranking? i feel like if you were looking back now, saying what was the best bball program of the 1980's, you would probably want to weigh every year equally. but, in terms of the best basketball program of the last ten years, I think people would mostly tend to weight the more recent seasons heavier. would you agree? or do you feel all-time dynasty rankings should be weighted for more current seasons as well?

Quote: Originally posted by metsmax on 2/16/2010countdown to post by Colonel Mustard explaining his systemm.......I think what you describe is cool - and my personal view is that I would not sweat the details - would be interesting although that is a ton of work!

thanks mets! well honestly, now that i have all the data imported, and knowing i will go through the effort of formatting the output and posting everything, programming in a more complex, better ranking system is a pretty small time investment compared to the whole. also, its my favorite part! i love stuff like this, trying to rank these teams as well as i can is something i enjoy every minute of :) the rest of it is work, but devising the strategy is awesome. and looking at the results :)

Billy, this is quite an ambitious project and there are so many steps involved it is hard for me to completely follow. I don't want to undervalue this effort in any way but is it possible for you to explain the concept in a non-mathmatical way? Is the goal to rank each conference at all three levels so we could see for example that the Little East in Wooden is #3 and the Little East in Tark is #7? Following that are you going to say for example that Framingham St. is #1 in Naismith and #6 in Smith?

To me, the best way to value/compare conferences across all worlds is to look at the RPI and # of NT bids prior to the start of the NT. My reasoning is that, in my mind, a conference that averages 5 NT bids per season with an average of 4 of those teams losing in the 1st or 2nd round is still a tougher conference than one that averages only 3 bids even if all three make the E8.

I look at it like this, St. John's University is either 6th or 7th on the all-time NCAA D1 win list, but for this decade have been a bottom feeder in the Big East. Despite them having so many wins where would you rank them as far as dynasties go, top 10? top 20? top 30?

It's tough to call, I guess I'm looking at things more along the lines of prestige while you're talking about overall dynasty. The more I think about it, the more I feel weighing things equally or giving the more recent years a slight advantage are about an equal way of determing the best dynasties.

hard for weena to follow?

that is hard for me to even fathom.

well weena, i was trying to focus only on the team by team dynasty rankings in this thread, and was going to tackle the conf dynasty rankings later. but, what i really am going for in terms of these dynasty lists overall, is to come out with a list of the top programs and conferences across worlds. anyway, here are the specific lists i wanted to come up with for dynasties, and ill try to avoid the math :)

1) team by team current dynasty list. this would say for example, in first place in d3 over the past 10 years is west connecticut state from the little east conference in tark, with a score of 640.7. in second place is howard payne from the ASWC in wooden with a score of 638.4.

this would effectively be a "top 100 current program" list, per division. not meant to be definitive, just for fun. and this really applies to all of the lists, including these dynasty lists, the all times streaks lists (most consec sweet 16 appearances, etc), and the list of global accomplishments that is yet to come (will have misc all time lists, like most championships per coach, most NT wins per coach, best NT win % by coach with more than say 20 wins). i want to make the lists long enough and with enough variety to allow a large pool of coaches to compete for spots, not just vets, so i am shooting for about 100 teams per list. i might even do 100 for the bigger sets of lists (like the 30 in the all time streaks lists) and a few hundred for the smaller set of lists (like there would only be 1 list per division for the current dynasty rankings. unless i did a couple lengths of time. but still, i think many of those teams would overlap).

anyway, new coaches still would have no chance but when i started i would have thought it was cool to see some of the great programs and accomplishments chronicled, i would have found it very interesting seeing what the ceiling was (the greatest accomplishments of all time, like rails's 8 of 9 or ORs d1 3peat). and i would have loved to dream of one day getting my name up there. or maybe to check if the conf mate who beats me down so badly is on there :) for coaches starting to compete at a high level, as you start to get somewhat close to making the list, i would hope it would be an exciting motivator - it definitely would be for me! same for those on the lists, trying to make it higher. and, i hope people would take pride in their accomplishments on these lists, which i feel would add to the overall experience. so, thats really the goal of all this, from a really broad standpoint :)

back to the dynasty lists themselves...
2) i want to make an all time, team by team dynasty list. so, you might have on the 10 year d2 all time list, 1st place, rails at w chester pa with a score of 850. and so on. so this isn't a list of a team's ranking in each world, like in your example, but a ranking of teams against each other, across worlds. WIS has some ways of comparing teams within a world, and some people do dynasty rankings for their division (which i have always found super interesting, which is my inspiration for this project). but as far as i know, there are no cross world lists of any kind.

3) i want to do a current conference dynasty list, similar to the team by team current dynasty list, just for conferences. i don't want to do a flat sum of the members, because for example, i think if a conf has 4 titles in a 10 year span. that is impressive. but, if they belong to 1 school, i think from a conf standpoint that is significantly less impressive than if 4 schools have 1 each. but, at its core, the conference dynasty list would be based on the rpi and post season success of the member schools, as well as world strength.

4) finally, i want to do an all time conference dynasty list. i think its probably obvious from the above 3 what i am going for on this one.


i think one of the most interesting concepts in all of this is world strength. it seems you cannot fairly compare accomplishments across worlds without weighting the competitiveness of the world. in your example, what if that conf who averages 5 bids a season is in a world with 70 human coaches, and the one with 3 elite 8 teams is in a world with 200 human coaches? i think that the 3 elite 8 team conference is easily more impressive, with that in mind. so, i think you are right that rpi and # of bids are good things to go off of for conference strength, base. but i feel it really needs a world strength factor to compare across worlds. because the base accomplishments, both for individual teams and conferences alike, are certainly of different levels of achievement when the competitiveness of the world varies significantly.

Quote: Originally posted by kmasonbx on 2/16/2010I look at it like this, St. John's University is either 6th or 7th on the all-time NCAA D1 win list, but for this decade have been a bottom feeder in the Big East. Despite them having so many wins where would you rank them as far as dynasties go, top 10? top 20? top 30? It's tough to call, I guess I'm looking at things more along the lines of prestige while you're talking about overall dynasty. The more I think about it, the more I feel weighing things equally or giving the more recent years a slight advantage are about an equal way of determing the best dynasties.

well, again, what do you think about current dynasties vs all time? with prestige, or the measure of the greatest teams of the day, i feel like there probably should be more weight on the more recent years. especially as at some point all the older years are worth 0 :)

but, with all time dynasties, i liken it to (sorry to reuse the example) trying to pick the best team of the 80s. does 89 really count for more than 80, assuming the tournament field did not expand? i feel like people would say ok, how many wins, championships, final 4s, tourney appearances, etc does each team have. but they wouldn't say well, because team A won in 80 and team B won in 89, then team B has an edge. i am doing something a little different, which is greatest dynasties of any time period. but its going to mostly be historical. so to answer "what is the best 10 year run of all time", i think all 10 years should be equal.

so overall, my feeling has basically become, i am pretty decided on equal weighting in all-time lists. and i am still open on the current lists. but i feel like a weighting is probably appropriate, although i don't want it to be too extreme.

Quote: Originally posted by oldave on 2/16/2010hard for weena to follow?that is hard for me to even fathom.

come on davey :) for a guy who managed to make so many of the all-time streak lists (including 3 second places - very nicely done!), i'm sure you can manage to figure out the lists themselves :P

seriously, doc, this is awesome... i think this kind of stuff is the real lifeblood of the game and i really dont think it would have the same effect if WIS did it themselves.

ive got a couple of comments here, but really, i think i would love what you are doing regardless of whether any of my input is used.

1) there have been a couple of comments along the lines of "dont sweat the small stuff". and i would tend to agree... although.... if you are able to get this thing set up to where it would auto-calc once you set the initial formula... then hey, why not sweat the small stuff? (just so long as its not that much sweat) heck, once you are done, just for kicks, compare what you have with lists that others have gnerated (I am assuming they used simpler models using only NTwins and rpi) with your final list and see if the little micro-adjusters caused much difference.

2) i tend to think season should be weighted equally on alltime list, though kman does raise valid points. i would tend to think you would be more likely to get it right for HD if you weighted them equally. plus, if you also have a current dynasty list, that is where those with more recent success will get thier due.

3) someone referred to "greatest program of the 80's" above... i know it would be way down the road from here, but i sure think it would be cool if you could develop a list of top dynasties for each 10season period for each world(div).

4) world ranking - this is an awesome idea! obviously not easy to come up with a ranking system that will be widely accepted, but if you can do it... that would really fire up some "world pride", wouldnt it? obviously even with a good system there will be arguments... but thats kind of the point,,, isnt it?

anyways, really looking forward to seeing the finished product(s).

and kinda hoping that this little project might distract you from gameplanning during the latter portion of Tark non-conf this season ;-)

i think you weight things on current lists, but not on all-time. the Wooden d3 all-time rankings would appear very very different from the current rankings i compile.