Creating Projections (1 Viewer)

[mattyice82]

I wanted to do my own projections this year and was wondering if anyone has any tips/suggestions on how to tackle this.Thanks!!!mj

 

[Angry Beavers]

Team by Team.. much easier than trying to go position by position....

 

[FF Ninja]

The most useful thing about the FBG projections is that they list the last 3 years of stats for each team, so I use them solely for this and ignore their actual projections. Those historical stats only help if the functional OC is intact. By functional OC, I mean if the head coach is really running the offense then ignore any changes to the OC - like when Kyle Shanahan was the OC for the Texans, but they were running Kubiak's offense.

If an offensive coach or OC moves to a new team, use his trends in his last job when projecting the new team. But keep in mind personnel. Good OCs will play off their team's strengths, not just strictly adhere to their own plan.

But keep in mind that this is largely an exercise in futility as far as results go. No one projected Welker to absolutely blow up last year and no one projected AP to hit 2000 yards this year or for Alfred Morris to get 335 carries. But I still think it is an important exercise. It makes you think about the situation so you don't just go projecting 1500 rushing yards and 12 TDs for a player when the previous year the entire team only rushed for 1200 yards and 4 TDs. I don't remember the real numbers, but I'm lightly referencing the Richardson hype. And when you're forced to project a QB's numbers, it keeps you honest on the WRs. Or might bring to light a situation where you start looking deeper and players like Andre Roberts or Brandon Myers suddenly jump onto your radar because somebody has to catch passes.

 

[Sweet Love]

Team by Team.. much easier than trying to go position by position....

Interesting...do you do a set of projections after the draft and then just assign individual stats based on training camp battles, etc.? I kind of like this method as you are less likely to get caught up by a guy who breaks out in the preseason, you assign him a ton of yards/stats, and then forget to take away from somewhere else (as he will likely be eating into someone's stats).

 

[FF Ninja]

Team by Team.. much easier than trying to go position by position....

Interesting...do you do a set of projections after the draft and then just assign individual stats based on training camp battles, etc.? I kind of like this method as you are less likely to get caught up by a guy who breaks out in the preseason, you assign him a ton of yards/stats, and then forget to take away from somewhere else (as he will likely be eating into someone's stats).

Interesting? Do people actually attempt to do full sets of projections player by player? Team by team is intuitively far and away the easiest/best way to do projections. All those crazy projections you see in player spotlights? Those are done by people who do individual projections. If they attempted to fit them into team projections there's no way they'd be able to rationalize the numbers they project. Team projections force people to consider how many targets it takes for guys like Roddy White to put up the numbers they do.

 

[Angry Beavers]

Team by Team.. much easier than trying to go position by position....

Interesting...do you do a set of projections after the draft and then just assign individual stats based on training camp battles, etc.? I kind of like this method as you are less likely to get caught up by a guy who breaks out in the preseason, you assign him a ton of yards/stats, and then forget to take away from somewhere else (as he will likely be eating into someone's stats).

Yes, after the draft and training camp is underway. I will look to see what the offense has done the last year or two, what changes were made to personnel, OC's, HC etc and who may be emerging, getting close to retirement etc. It is to some extent an exercise in futility as you are clearly going to miss things, but I think the advantage is 1) it forces you to understand each team and what went into the number. There is a difference between projecting 800 yards and 5 td's as a reasonable expectation for a breakout candidate ( who may not quite reach that goal and is just as likely to explode past it) or what you can reasonably expect from a veteran but likely not more and 2) keeps you honest as you can't predict two 1000 yard rb's from a team without a special situation.

 

[by_the_sea_wannabe]

Team by Team.. much easier than trying to go position by position....

Interesting...do you do a set of projections after the draft and then just assign individual stats based on training camp battles, etc.? I kind of like this method as you are less likely to get caught up by a guy who breaks out in the preseason, you assign him a ton of yards/stats, and then forget to take away from somewhere else (as he will likely be eating into someone's stats).

Interesting? Do people actually attempt to do full sets of projections player by player? Team by team is intuitively far and away the easiest/best way to do projections. All those crazy projections you see in player spotlights? Those are done by people who do individual projections. If they attempted to fit them into team projections there's no way they'd be able to rationalize the numbers they project. Team projections force people to consider how many targets it takes for guys like Roddy White to put up the numbers they do.

I do individual and then see if the the wr/rb combos are within reason with the qb. No need for the numbers to be exact.

 

[Gawain]

The most success I've ever had was when I stopped making my own projections and started slotting players into already compiled stats.I average the past 3 years of data by finish and position. This gives me a stat out put before I even have a player to go along with it. (WR 20 over the past 3 years averaged 70 catches 975 yards and 7.5 TD's for example)Once I have that data, I rank players. Then it's as easy as matching up the production list with the player list.

 

[Please See Mine]

The most useful thing about the FBG projections is that they list the last 3 years of stats for each team, so I use them solely for this and ignore their actual projections. Those historical stats only help if the functional OC is intact. By functional OC, I mean if the head coach is really running the offense then ignore any changes to the OC - like when Kyle Shanahan was the OC for the Texans, but they were running Kubiak's offense.If an offensive coach or OC moves to a new team, use his trends in his last job when projecting the new team. But keep in mind personnel. Good OCs will play off their team's strengths, not just strictly adhere to their own plan.

Good point.I was thinking about a model for more week-to-week projections and trying to incorporate the health/quality of a team's starters (particularly offensive line). I think that getting a good read on the O-Line may be undervalued in trying to determine potential production.

 

[mlball77]

Team by Team.. much easier than trying to go position by position....

Interesting...do you do a set of projections after the draft and then just assign individual stats based on training camp battles, etc.? I kind of like this method as you are less likely to get caught up by a guy who breaks out in the preseason, you assign him a ton of yards/stats, and then forget to take away from somewhere else (as he will likely be eating into someone's stats).

Interesting? Do people actually attempt to do full sets of projections player by player? Team by team is intuitively far and away the easiest/best way to do projections. All those crazy projections you see in player spotlights? Those are done by people who do individual projections. If they attempted to fit them into team projections there's no way they'd be able to rationalize the numbers they project. Team projections force people to consider how many targets it takes for guys like Roddy White to put up the numbers they do.

I seem to do full player projections every 2-3 years. I enjoy it, but it's a lot of work, yet it makes me think through my stances on player much more deeply. I approach it like you do, team by team. It's really interesting to see how each player fits into the puzzle for their team... and this approach keeps things much more realistic than doing projections by positions or whatever, imo. The end numbers are nice/fun, but the process is what I find much more valuable. I learn a lot.

 

[butcher boy]

I used to have a spreadsheet where I would do a first pass by player, then the spreadsheet would add up the player stats by team and compare with last year's team totals. I'd adjust from there based on whether I thought teams would be better or worse on offense.

 

[Bayhawks]

The most useful thing about the FBG projections is that they list the last 3 years of stats for each team, so I use them solely for this and ignore their actual projections. Those historical stats only help if the functional OC is intact. By functional OC, I mean if the head coach is really running the offense then ignore any changes to the OC - like when Kyle Shanahan was the OC for the Texans, but they were running Kubiak's offense.

If an offensive coach or OC moves to a new team, use his trends in his last job when projecting the new team. But keep in mind personnel. Good OCs will play off their team's strengths, not just strictly adhere to their own plan.

But keep in mind that this is largely an exercise in futility as far as results go. No one projected Welker to absolutely blow up last year and no one projected AP to hit 2000 yards this year or for Alfred Morris to get 335 carries. But I still think it is an important exercise. It makes you think about the situation so you don't just go projecting 1500 rushing yards and 12 TDs for a player when the previous year the entire team only rushed for 1200 yards and 4 TDs. I don't remember the real numbers, but I'm lightly referencing the Richardson hype. And when you're forced to project a QB's numbers, it keeps you honest on the WRs. Or might bring to light a situation where you start looking deeper and players like Andre Roberts or Brandon Myers suddenly jump onto your radar because somebody has to catch passes.

I don't necessarily agree with all your opinions/conclusions that you draw, but this is excellent posting, right here.

 

[GregR]

The most success I've ever had was when I stopped making my own projections and started slotting players into already compiled stats.I average the past 3 years of data by finish and position. This gives me a stat out put before I even have a player to go along with it. (WR 20 over the past 3 years averaged 70 catches 975 yards and 7.5 TD's for example)Once I have that data, I rank players. Then it's as easy as matching up the production list with the player list.

This is often referred to as AVT, or Average Value Theory.One of the biggest issues it has is that if you use those projections you tend to over project the top players because you don't take into account the odds that the player you chose will not end up in that spot.As an example, I may think the odds are Rodgers is QB1 30% of the time, QB2 25%, QB3 10%, QB4 10%, etc. If I just give him the points for QB1 then if I use those points to determine his value and slot a draft position or set an auction salary for him, I'm instead saying I think he is QB1 100% of the time. Since I don't feel that way, his value should be lower, accounting for my belief he may not be QB1.Put another way with a weekly scoring example, if you ask me for week 1 how many yards will QB1 of that week throw for, I should probably say around 400 or whatever the average is for the top QB each week. But if you ask me how many yards do I think Aaron Rodgers will throw for, since his career average is closer to 250, I should say that (adjusting for matchup of course). Even if I think he has the best chance to be QB1 in week 1, that doesn't mean the best answer is just assigning him QB1 points for that week.

 

[cvnpoka]

I wanted to do my own projections this year and was wondering if anyone has any tips/suggestions on how to tackle this.Thanks!!!mj

1. predict how many plays a team will run2. predict run pass ratio3. predict qbs stats4. predict breakdown of carries and predict rb stats5. predict breakdown of targets based on plays and qb prediction6. predict wr and te an rb passing statsits pretty tedious as that is obv glossing over a ton of details.

 

[Maurile Tremblay]

I wanted to do my own projections this year and was wondering if anyone has any tips/suggestions on how to tackle this.Thanks!!!mj

1. predict how many plays a team will run2. predict run pass ratio3. predict qbs stats4. predict breakdown of carries and predict rb stats5. predict breakdown of targets based on plays and qb prediction6. predict wr and te an rb passing statsits pretty tedious as that is obv glossing over a ton of details.

That's somewhat similar to how I do it.For my weekly projections, I start by projecting how many points the team's offense will score (using the Vegas spread and over/under, and taking into account and subtracting likely defensive/special teams points and field goals). Based on the expected offensive points and the expected margin of victory or defeat, I project the team's number of offensive plays and the run-pass ratio. Once I have that, I've got the team-wide rush attempts and pass attempts (and targets, which will equal pass attempts). From there, it's a matter of distributing the rush attempts and targets to the individual players, projecting each player's expected completions per pass attempt, yards per pass attempt, touchdowns and interceptions per pass attempt, yards per rush, touchdowns per rush, receptions per target, yards per reception, and touchdowns per reception. Then I make whatever adjustments are needed to see that the passing stats match up with the receiving stats, and that the overall TDs match up with the team's expected offensive points.For my preseason projections, I kind of do the same thing, generally speaking, but everything is multiplied by 16.

 

[mattyice82]

thanks for all of the feedback, much appreciated!!!!

 

[Dinsy Ejotuz]

For my weekly projections, I start by projecting how many points the team's offense will score (using the Vegas spread and over/under, and taking into account and subtracting likely defensive/special teams points and field goals). Based on the expected offensive points and the expected margin of victory or defeat, I project the team's number of offensive plays and the run-pass ratio. Once I have that, I've got the team-wide rush attempts and pass attempts (and targets, which will equal pass attempts). From there, it's a matter of distributing the rush attempts and targets to the individual players, projecting each player's expected completions per pass attempt, yards per pass attempt, touchdowns and interceptions per pass attempt, yards per rush, touchdowns per rush, receptions per target, yards per reception, and touchdowns per reception. Then I make whatever adjustments are needed to see that the passing stats match up with the receiving stats, and that the overall TDs match up with the team's expected offensive points.

Pretty interesting MT. I assumed you all started with his David's game predictor and worked down from that. Does that mean your (and Bloom's) weekly projections are completely independent of David's?

 

[Maurile Tremblay]

For my weekly projections, I start by projecting how many points the team's offense will score (using the Vegas spread and over/under, and taking into account and subtracting likely defensive/special teams points and field goals). Based on the expected offensive points and the expected margin of victory or defeat, I project the team's number of offensive plays and the run-pass ratio. Once I have that, I've got the team-wide rush attempts and pass attempts (and targets, which will equal pass attempts). From there, it's a matter of distributing the rush attempts and targets to the individual players, projecting each player's expected completions per pass attempt, yards per pass attempt, touchdowns and interceptions per pass attempt, yards per rush, touchdowns per rush, receptions per target, yards per reception, and touchdowns per reception. Then I make whatever adjustments are needed to see that the passing stats match up with the receiving stats, and that the overall TDs match up with the team's expected offensive points.

Pretty interesting MT. I assumed you all started with his David's game predictor and worked down from that. Does that mean your (and Bloom's) weekly projections are completely independent of David's?

I haven't really discussed with David or Sigmund exactly how they do their projections, so I don't know how different our methods are. I don't use the Game Predictor at all. It may seem weird that we haven't really brainstormed collectively about it. But as much as I'd like to trade thoughts and start copying any good ideas they have that I haven't come up with on my own, I do think there's value in keeping our approaches completely independent. I think three distinct views, coming from different angles, are more valuable than one predominant view with a couple minor variations.

 

[Dinsy Ejotuz]

as much as I'd like to trade thoughts and start copying any good ideas they have that I haven't come up with on my own, I do think there's value in keeping our approaches completely independent. I think three distinct views, coming from different angles, are more valuable than one predominant view with a couple minor variations.

Totally agree -- it's a feature not a bug. Just didn't realize you all worked independently.