Posted by ethan66 on 4/14/2011 7:36:00 PM (view original):

Posted by professor17 on 4/14/2011 6:26:00 PM (view original):
Your opponents' winning percentage is far more important than your opponents' RPI, when it comes to getting a strong RPI out of your non-conf schedule. A lot of people mistakenly think that scheduling a middling or bottom team from a really strong conference will help their RPI, but I'd much prefer to play a sim (or human, for that matter) team that will win 20+ games against weak opposition. RPI is 25% your winning percentage, 50%  your opponents' winning percentage and 25% your opponents opponents winning percentage. So how many games your opponent will win is twice as important as how tough their schedule is.

Thanks for all the input, you guys.  You've given me a lot to think about as I look ahead to this next season.

Professor17, when you say you prefer to play a team that will win 20+ games, how do you determine that?  Digging through their past season's schedule and looking at returning players?

How much do you all regard a potential opponents' prestige when scheduling?  Does it help RPI or is it a non-factor?
 

Ideally I look for teams that went at least .500 in their conference, return 9-10+ players, and are going to play the vast majority of their games against sim teams. 

Posted by jdno on 4/14/2011 9:13:00 PM (view original):
Professor, good example, thanks.  But shouldn't we be multiplying the OOWP by 25% instead of SOS by 25% as I think you've done?  Getting the right OOWP is what would be tedious.  And then I can't remember if the home and away multipliers factor in for OWP and OOWP.  I know it was discussed recently on the board, but I already forget what the conclusion was. 

Yeah, I was wondering that as I was writing the example, because I wasn't completely sure whether SOS was a formula calculated number, or just an aggregate winning percentage. But either way, I think the conclusion is the same... the gap between the two teams would just narrow. It's not just that OWP is worth twice as much as OOWP in the RPI formula. It's also that OWP is going to be a higher absolute number anyway if you schedule properly, simply because OOWP is going to aggregate a lot more teams with wider variability in winning pct... it's going to regress towards the average quicker than OWP. If you follow what I'm saying...

I do get what you're saying and agree with the latter part of your argument.  I just wish I could accurately calculate the numbers in a case like you put forth to see how big an advantage it is to schedule a decent (but beatable) 20-win SIM over a .500 Big 6 kind of school.

It depends on 2 things, how good your team is and how good your conference is. If you have a great team it really doesn't matter, schedule as hard as possible. However if your like the vast majority of teams and your team isn't going to be great and your in an average conference you want to schedule as hard as you can while putting together a schedule that will get you 7+ wins. So if you think you can win 7 games against a top 50 SOS do that, if you think you can only get 7 wins if you put together a top 100 SOS then do that.

Now if you're going to be a solid team but play in a great conference I'd play a very soft schedule filled with road games whether I'm playing sims I expect to be good (check their current season record and how many returning players they have, a .500 team with 10 returning playesr is bound to be a .600 team the next season) and users in sim filled conferences that's the road I go. For an example of this check out my schedule with West Virginia, I scheduled 8 sim AI's all on the road. The reason I do this is because my conference is so good that it doesn't matter how bad my NC schedule is the conference schedule is going to pick up my SOS to a respectable level. But be sure to not just schedule any sim, make sure they are in sim filled conferences and have a chance to be decent so they end up winning 13+ games.

I also always play at least 6 road games.

Could you extract your opponent's opponent's winning percentage by working backwards from your own record? That makes up 25% of the total.  Your opponent's record you could get from the schedule...at 50%of the total- once you know two of the three inputs couldn't you treat it sort of like a "Solve for x" problem?

 

Of course, I could just be not thinking about it right...

Another trick is to arrange your schedule so that it goes from least difficult to most difficult.  This is helpful if you think you will have to play a freshman or two and they will improve rapidly from game 1 to game 10.  You don't get any more credit for beating a cupcake in game 10 than you do in game 1, so give your youngsters the cupcake to eat when they need that light snack the most!

Another idea is that if you are going to schedule death matches (teams that you think are much better than you and ou have little chance to beat) then you might as well play them on the road.  If you are going to probably lose, then get the most value for your loss, since the road game will help your RPI more.  Similarly, if you play a team that you think you have an overwhelming advantage over, play them on the road, too.  The teams you want to play at home are the teams that are closer to you in ablility, where your home court advantage will help you the most in determining the outcome of the game.

One more thing:  there are easy games that give you a good chance to win. It probably won't hurt to give yourself 3 or so of those. Then there are games that are a complete waste of time.  Under no circumstances do you want teams with 300+ RPI on your schedule. But, a 225 RPI here and 250 RPI there, aren't a bad thing in moderation (unless you are aiming for the top 10).  This doesn't apply if you are trying to go all out, but if you are in the position that you feel you could use a few breaks, just make sure you don't schedule any wastes.  If you are a team that is borderline top 50 or borderline top 75, you will benefit from going 3-0 against RPI 200-250.  Just don't lose! lol

As a practical consideration, if you are going to micromanage your schedule, then you probably want to get it done fairly early in the season. This is simply because (a) it is easier to get games earlier than later (b) if you are like me then you usually say yes to offers, so rather than having to reject ones that don't fit your plan, get the schedule done.

ryandaniel, yes, this is basically how you'd have to do it if doing it by hand.  Of course, your OWP should have the home/road multipliers applied to it, which makes it quite tedious.

However, for our purposes, a simple example ignoring the home/road factors for OWP and OOWP should be close enough.  So, with this in mind, I looked at some Rupp data I have access to.

I chose E. Michigan, 20-8, RPI 135, SOS 306, RPI value = 0.5152, SOS value = 0.4464 as the 20-win SIM with the cupcake schedule
I also chose Arkansas (16-17record, playing in the ultra-dominant SEC ) as the 0.500 Big 6 team w/a human coach RPI 58, SOS 7, RPI = 0.5769, SOS = 0.6101

I made use of the relationships:

Eq. 1:  RPI = 0.25*WP + 0.50*OWP + 0.25*OOWP, which, after some algebraic  manipulation can be expressed as:

Eq. 2:  RPI = 0.25*WP + 0.75*SOS

Since HD calculates RPI and SOS for us, we can easily calculate WP for any given team and this would take into account the home/road factors for a given team's WP.

So, for E. Michigan, we have Eq. 2 as 0.5152 = 0.25*WP + 0.75*0.4464, solving for WP gives WP = 0.7216.  Next, I calculated OWP by adding up the wins and losses for E. Michigan's opponents from its Schedule page.  Technically one should account for home/road wins/losses here, but I ignored this part for simplicity.  EMU's opponents' aggregate win/loss record was 325 - 455, for an OWP of 0.4167.  Now we can solve for OOWP using Eq. 1.  Doing this, we find OOWP = 0.5058

In sum, for E. Michigan, we have WP = 0.7216, OWP = 0.4167, OOWP = 0.5058

Doing a similar exercise for Arkansas, we find WP = 0.4773, OWP = 0.5671, OOWP = 0.6961

So, if you're deciding to play E. Michigan or Arkansas, which benefits your RPI more (assuming you have an equal chance of beating each one)?  Let's look at Eq. 1 again and we'll ignore the 0.25*WP factor in that equation since we're assuming you'd either win or lose vs. both of them, so that's a wash.  Notice we'll have to "slide" WP, OWP, and OOWP each to the right b/c now we're looking at the impact of playing each of these teams on OUR team.  We're not looking at things from EMU's or Arkansas's standpoint anymore.

For EMU, RPI impact = 0.5*OWP + 0.25*OOWP = 0.5*(0.7216) + 0.25*(0.4167) = 0.4650
For Arkansas, RPI impact = 0.5*(0.4773) + 0.25*(0.5671) = 0.3804

So, yes, as we kind of suspected, as far as your RPI goes, it's better to play the SIM with the soft schedule than a mediocre Big-6 school, but this assumes you can beat both teams, which may not be the case.  However, I think looking at the formulas, if you lose to ARkansas but beat EMU, you end up hurting yourself even more because the first factor in the RPI equation is 0.25*WP.  If you lose to Arkansas, your WP will be lower right off the bat vs. beating EMU. 

Also, we have to account for the fact that this is just 1 game out of ~30 that you'll play in the season.  It's not like you play EMU all 30 games or Arkansas all 30 games.  

Lastly, notice that Arkansas's meatgrinder-like OOWP of 0.6961 doesn't affect your RPI at all in this exercise.  The SEC is the deepest and most difficult conference, but this factor gets lost in the RPI impact for YOUR team in the calculations. 

Anyhow, a thoughtful mathematical exercise here for sure.  Thanks guys for the impetus to get me off my a$$ and work through the numbers after always wondering about this topic.

I may have messed up in my calculations or logic here, please point it out if you think this is the case.

^ So, in short: your opponents' winning percentage is twice as important as who they played. 

If RPI and SOS generally converge as the season progresses as mully said, then yes, jkline, it certainly seems that way, tweaked for the home/loss multipliers.

The other basic equation that I didn't show above but is used in deriving Eq. 2 is simply:

SOS = 2/3*OWP + 1/3*OOWP, which essentially states your point.

I use the lazy method - I don't schedule. I do accept all invites though. As a very very broad rule of thumb this results in my bad teams having weak schedules and my good teams having better schedules. 

when i first took over shepard i scheduled as many easy teams as possible to rack up wins... i want 21+ wins and needed 10 OOC to make it there... now that my team has some talent i'm going for much harder schedules... and as many games on the road as possible. I saw teams with top 50 RPIs with only 14-15 wins whereas I had 22 wins and a 80 RPI... so now i'm working toward the middle.

Posted by jdno on 4/14/2011 11:45:00 PM (view original):
ryandaniel, yes, this is basically how you'd have to do it if doing it by hand.  Of course, your OWP should have the home/road multipliers applied to it, which makes it quite tedious.

However, for our purposes, a simple example ignoring the home/road factors for OWP and OOWP should be close enough.  So, with this in mind, I looked at some Rupp data I have access to.

I chose E. Michigan, 20-8, RPI 135, SOS 306, RPI value = 0.5152, SOS value = 0.4464 as the 20-win SIM with the cupcake schedule
I also chose Arkansas (16-17record, playing in the ultra-dominant SEC ) as the 0.500 Big 6 team w/a human coach RPI 58, SOS 7, RPI = 0.5769, SOS = 0.6101

I made use of the relationships:

Eq. 1:  RPI = 0.25*WP + 0.50*OWP + 0.25*OOWP, which, after some algebraic  manipulation can be expressed as:

Eq. 2:  RPI = 0.25*WP + 0.75*SOS

Since HD calculates RPI and SOS for us, we can easily calculate WP for any given team and this would take into account the home/road factors for a given team's WP.

So, for E. Michigan, we have Eq. 2 as 0.5152 = 0.25*WP + 0.75*0.4464, solving for WP gives WP = 0.7216.  Next, I calculated OWP by adding up the wins and losses for E. Michigan's opponents from its Schedule page.  Technically one should account for home/road wins/losses here, but I ignored this part for simplicity.  EMU's opponents' aggregate win/loss record was 325 - 455, for an OWP of 0.4167.  Now we can solve for OOWP using Eq. 1.  Doing this, we find OOWP = 0.5058

In sum, for E. Michigan, we have WP = 0.7216, OWP = 0.4167, OOWP = 0.5058

Doing a similar exercise for Arkansas, we find WP = 0.4773, OWP = 0.5671, OOWP = 0.6961

So, if you're deciding to play E. Michigan or Arkansas, which benefits your RPI more (assuming you have an equal chance of beating each one)?  Let's look at Eq. 1 again and we'll ignore the 0.25*WP factor in that equation since we're assuming you'd either win or lose vs. both of them, so that's a wash.  Notice we'll have to "slide" WP, OWP, and OOWP each to the right b/c now we're looking at the impact of playing each of these teams on OUR team.  We're not looking at things from EMU's or Arkansas's standpoint anymore.

For EMU, RPI impact = 0.5*OWP + 0.25*OOWP = 0.5*(0.7216) + 0.25*(0.4167) = 0.4650
For Arkansas, RPI impact = 0.5*(0.4773) + 0.25*(0.5671) = 0.3804

So, yes, as we kind of suspected, as far as your RPI goes, it's better to play the SIM with the soft schedule than a mediocre Big-6 school, but this assumes you can beat both teams, which may not be the case.  However, I think looking at the formulas, if you lose to ARkansas but beat EMU, you end up hurting yourself even more because the first factor in the RPI equation is 0.25*WP.  If you lose to Arkansas, your WP will be lower right off the bat vs. beating EMU. 

Also, we have to account for the fact that this is just 1 game out of ~30 that you'll play in the season.  It's not like you play EMU all 30 games or Arkansas all 30 games.  

Lastly, notice that Arkansas's meatgrinder-like OOWP of 0.6961 doesn't affect your RPI at all in this exercise.  The SEC is the deepest and most difficult conference, but this factor gets lost in the RPI impact for YOUR team in the calculations. 

Anyhow, a thoughtful mathematical exercise here for sure.  Thanks guys for the impetus to get me off my a$$ and work through the numbers after always wondering about this topic.

I may have messed up in my calculations or logic here, please point it out if you think this is the case.

One thing I can think of is that when scheduling the 'easy sims that get wins' you ARE likely assuming a higher risk of any given team doing a total flamout than you are with a human?

Yeah, that's definitely a risk, esp. if other good teams are on their NC schedule as well.  In practice, I rarely schedule a "good" sim, I usually schedule a good human-coached team in a lesser  conference, figuring they'll coach better (to minimize the exact risk you're talking about) and will schedule better as well.  A NC schedule of 20-win human coached teams is my target if I think my team is good enough to win at least 6 of them.  This gets adjusted a bit higher or lower depending on the quality of teams within my conf.