SGP valuation method

[rotodog]

Todd,

I know it requires more than a couple sentences, but you explained to me why SGP valuation was flawed once. Once you said it, I remembered it made immediate sense to me. Although I have come to the conclusion exact valuation is an exercise in futility, can you explain in a couple sentences why it was the case. I have been away from serious Roto for a couple years and for the life of me, i cant remember... Thanks

[Todd Zola]

FWIW, the essay explaining this is available in the Platinum archives and at some point, hopefully sooner than later we will be taking this stuff out from behind a firewall, but for now, players are valued based on how many points they "gain" in the standings, based on "average" standings. Let's say the average number of HRs between standings places is 6 -- if a player hits 18 HR, he gets 3 SGP (3x6=18). This is too simplistic, but these SGP are added across all the categories, summed and value is given based on the percentage of SGP each player has with respect to the total pool.

The flaw is no consideration is given for the baseline level of each stat (last place total) before you begin accruing points (1 pt to 2 pts, etc). The distribution between categories is not equal, mainly SB in hitting and SV is pitching. Due to the singular nature of those stats, some teams fare poorly (or purposely bag) SB and/or saves. So some of a SB guys value in terms of his SGP in effect go towards eating away at the baseline for HR, RBI etc and not towards his SB value.

The way this is "visualized" is after assigning SGPs, in theory, in a 12-team league, there should be 66 (12th to 11th = 1, 12th to 10th =2...12th to 1st = 11, etc.) But in practice there are somewhere in the 100s, and each category is different.

[rotodog]

Thanks Todd....Now I remember....

One question? If it is this obvious to me when you explain it why do certain Bigtime ROTO websites use SGP to value players for their own valuations and valuation tools? For simplicity? Because they sell Mcgee's Book? Just wondering actually..

[Todd Zola]

Sorry, I can't speak for other sites.

But as an FYI -- they no longer use SGP.

[viper]

I like to visualize SGP differently. I believe there is a flaw looking at the last place team -and any team at either end of the category scores. Category dumping occurs and this artificially lowers the totals for the last place team. I prefer looking at the average score in a category.

For example, assume the average HR total in an NFBC league is 280 and the delta for HRs is 5 [both close to actual]. This means that an average player accumulates 20 HRs. A player with 30 HRs gets 2 SGP points of value. A player with 10 HRs has -2 SGP points. Once you get over the fact that some players are plus and others are negative, this works out. The average SB total for a team is about 140 or 10 per player. The delta for steals is also close to 5. A player with 20/10 is average. A player with 25/5 or 15/15 will also have the same SGP point value - they are all zero. If you end up with cumulative zero SGP points for a category, you end up with the average total of, in these cases, 280 or 140. If your team has a positive SGP total, you are above average. By keeping a track of your individual category totals, you can see where you can expect an above average catgory point total and where it will be below average. The math works out. It should be noted that I also used the McGee regression analysis but I didn't use it for the 15 thru 1 category totals but for a 13 thru 5. My reasoning is that the extreme scores are not "trustable" as some people overkill a category and other dump. I trust this more in the no-trade NFBC league but I think it hold in all leagues. Even if there is no dumping the delta is close to the same in the 13-5 range or 10-4 in a 12-team league.

[Todd Zola]

FWIW, over the years (and as recently as last night), Gary and I have discussed valuation variations. Something we are considering messing around with is VERY similar to viper's methods, but using the category target you want to hit as a baseline and adjusting value based on how each player helps you attain that target, with respect to what is still available in the pool.

As a overly simplistic example, say you need 20 SB to reach your SB target and you have reached the others already. This method would put the guys that project for the most steals at the top of the "value" list, even if someone else were more valuable in an overall sense. Let's say Rajai Davis and Marcus Thames were available, even though Thames has more "raw value", this adjusted system would have Davis with more value for your team.

[rotodog]

FWIW, over the years (and as recently as last night), Gary and I have discussed valuation variations. Something we are considering messing around with is VERY similar to viper's methods, but using the category target you want to hit as a baseline and adjusting value based on how each player helps you attain that target, with respect to what is still available in the pool.

As a overly simplistic example, say you need 20 SB to reach your SB target and you have reached the others already. This method would put the guys that project for the most steals at the top of the "value" list, even if someone else were more valuable in an overall sense. Let's say Rajai Davis and Marcus Thames were available, even though Thames has more "raw value", this adjusted system would have Davis with more value for your team.

Todd, I referenced this a bit in Vipers thread about drafting toward targets and the flaw it presents as a stand alone. But If I were going to use a tool like this, I would want it to adjust negatively also. For example, If I have A team with Low AVG Sluggers and need avg, would it apply a little more negative weighting to a Mike Cameron type in the AVG category on the fly? His AVG will already have some sort of Negative value in the initial valuation, but will it have more Neg value if you need BA to hit a target?

The same holds true if you have a Team of HIGH AVG types and Conor JAckson comes up. (FYI, He is always undervalued to begin with). With 15 HR power at and 80 Run , 80 RBI his value as a 1b comes in the form of his .300 avg and a few steals. But If you need power to hit a target, would Conor's value on the fly come down because you can absorb a .250 hitter and not affect your AVG target?

This make any sense to you?

[rotodog]

Sorry, I can't speak for other sites.

But as an FYI -- they no longer use SGP.

Thanks..
I told you I have been away for a couple years