History repeats itself; whether it’s relating to the prepayment habits of an interest rate sensitive consumer, the behavioral aspects of the stock market, or individual baseball player statistics …history repeats itself!

Here’s the concept to understand: Why does history repeat itself? If you can answer this question, regardless of what you’re analyzing, you’ll be one step ahead of everyone else. The key word here is:  Limitations.

Why does history repeat itself when accessing a baseball player’s ability? Yep you guessed it: Limitations! There are limitations in a ball player’s ability and they come in two forms: physical and mental. Alex Sanchez is never going to hit 50 homeruns in a MLB season, even if he IV’s a direct line of horse growth hormones for 3 years (and apparently he tried some form of steroids). It’s just not within his physical structure. Sanchez has a certain range of offensive abilities that are quantifiable and forecast-able within a certain degree of accuracy. On the other end of this spectrum is a player like Craig Biggio who for years was able to synergize his mental & physical attributes enabling himself among the fantasy elite. Unfortunately, players similar to  Biggio's body type are also the type for fade off fast as their body's age. While a player like Mike Sweeney can still mash a ball 500 feet with a wretched back, a player like Biggio needs 100% of his health.  

We know, again within a certain degree of accuracy, that batters do not hit a peak until their 28^th birthday, some sooner, some later, but within a bell shaped range we know what to reasonably expect given the limitations of each player...barring an infusion of steroids of course....BTW: Look for overall offensive numbers to continue to decline thanks to the new testing in place.

Coming out of the minors we know that there will be a discrepancy in a player’s first year stats, verses his minor league stats. However, based on his past performances we can reasonably access who has and doesn’t have the potential to post a decent WHIP or strikeout total. It’s not exact, but it’s reasonable and it relies on history repeating itself.

As important as it is to follow history in an up trending pattern during a player’s career and conversely down trending during their latter seasons, it’s also important to throw out the anomaly seasons. This is just part of what our player projections algorithm is programmed to do.

Armed with a collection of Baseball data since 1880 (125 years), we use a hybrid non-linear regression formula to backward integrate the data. From there we basically use our statistics program (MatLab) to find the most important variables that have any statistical relevance from year to year. Using the end result of variables (moving trend analysis being the most weighed) we are able to produce results that are within acceptable ranges in predicting a player’s forward season.

From that point we scrutinize each player's computed projection and make what I call “subjective injections”.  Factors such as playing time are always subjective, as are human elements such as incentive, which obviously can not be factored by a machine, thus we intercede subjective analysis on at least 25% of the players (100% related to playing time expectations). 

Below is a series of charts showing how we stack up against the notion that projecting Baseball statistics is guess work. It is certainly not, especially when handled correctly by a statistician. Here's a graphical look comparing how we've done since our inception in 1999: 

The Chart above (to the right) depicts the reliability of using previous year's statistics to predict future year statistics for the top 50 batters. As you can see history is a good indicator, however the mean is within 50-70% within the 15-35% tolerance levels. The Chart to the right (above) shows the value that Fantistics adds to the predictability of baseball statistics, our 7 year mean lies between 65-90% accuracy between the tolerances, an incremental gain of exceeding 40% on average. Although pitchers are historically a very difficult group to predict or prognosticate. Fantistics has been remarkably accurate in predicting the trend for starting and relief pitchers over the last 6 seasons.

  Fantistics has been remarkably accurate in predicting the direction (up trending or down trending)  on 70-82% of position players, and 65 to 76% of Pitchers since our inception (see below)

As a side note, reliable forecasting would be extremely difficult if it were not for the large number of games played during a baseball season. A full season of 162 games provides us with a whole lot of information to make a reliable forecast based on the supple sample size.

Player Projected Stats are developed from the "Bottom -Up"

It is also very important to understand that the Projected Players Values and stats are going to add up to more than the possible year end results. Projecting MLB player stats or fantasy values based on the aggregate results and working them down to individual performances (the Top-Down approach) is not justified statistically. We have found through our research that the only legitimate way to forecast individual performances is through a bottom-up approach, since we are following/forecasting players and not teams or leagues in the typical fantasy league.

Forecasting from a "top-down" approach leads to a serious shortcoming in the prognostication of lesser known players and/or players whose roles are yet to be defined. The immediate concerns during the preseason are that roles (playing time) are not clearly defined among the new and the aging players. With this said the most dominate reason that our full projected stats and fantasy values will never add up to the year end aggregate results is because of Injuries and changes in playing time situations during the season. There is no way to avoid it, players get hurt, aside from those who have a history of injuries, there is no way to forecast which currently healthy players are going to get hurt.  Sure we forecast decreased playing time among those who have shown an inability to stay healthy, but a sizable portion of those players who land on the DL every year are 1st timers. Thus when we forecast our projections we have to rely on the known, forecasting from the bottom up (players verses aggregate totals), if we didn't we'd be forecasting between 75 - 80% of a player's potential stats....and we'd also have to leave off those players that will split time between the minors and majors this season.....A definite No-No. 

In closing, I know that many of you have read a multitude of different theories on projections, but I assure you that our logic is fundamentally sound and based on empirical evidence. Surely we do not possess a crystal ball, but having the right foundation is a key component to success on draft day....Gentlemen start your engines!   

- Anthony A. Perri

Article from previous seasons, but explains the VAM concept

VAM..Value Above Mean- It's difficult to say that any one concept is fully encompassing, but when it comes to Fantasy Drafting, the Fantistics' theory of Value Above Mean (VAM), is the definitive concept to winning on draft day. 

VAM: Value Above Mean (or average) is a computation which first measures a player's fantasy "worth" within his position and then secondly compares that "worth" to other players at different positions. VAM is also popularly known as the Value Based Drafting Method or the Position Scarcity theory. By taking the average fantasy worth for the typical number of starters at each position (based on your league specifics) we form a "baseline" which we can use to relatively compare each of the players at the particular position by. In a perfect world, we compared this strategy against "simulation owners" who were set to draft according to total fantasy points or $ valuations, and found VAM to be a considerably more effective tool. VAM is most effective in the early rounds of a draft (as evidenced by the graph below)

In 1998, we developed the concept using a statistical software package (MATLAB), and since then we have run over 100,000 simulations. 2012 will be our 14th year in which we have recommended this strategy to our subscribers.

The reason why VAM is effective is due to a theory called "positional value", or "position scarcity". The positional value system works because it values players in relation to their positional peers. In recent seasons, some fantasy folks have incorporated the position scarcity concept in their bidding or valuations without the numerical algorithm that we have quantified. Who can forget John Hunt's 2000 Baseball Weekly article on the "$50 Pedro Martinez bid", it was a brilliant move and opened some eyes to the true valuation of quality players at weak positions. A then healthy Pedro was so dominating in many major Rotisserie categories and essentially dwarfed the average pitcher available from the AL. Would Martinez have held as much clout if the average draft-able pitcher was the equivalent of a "Joe Mays"? Ask yourself this question: What is the drop-off from the best to the worst player that will be selected at each position in your draft? If your going from "Johan Santana" to "Eric Milton" at the pitchers position, and from "Todd Helton" to "Sean Casey" at the 1st base position...I'd take my chances with drafting the bottom of the barrel 1st Baseman than the rear in the pitching department. Couple this with the fact that bargains exist late in the draft, and you understand the importance of drafting quality players at notoriously weak positions. Hence this is the fundamental theory behind the VAM (Value Above Mean) concept. You want to elevate the value the top performers which come from the weakest positions, according to your draft parameters. This is the principal behind our draft software. Our algorithm will produce a value (a rank per se) for each player based on your league criteria. It's really relative to your league parameters, as an example, if your league requires a high number of outfielders, it's quite possible that outfielders will be in demand, creating a righteous artificial increase in their absolute value. 

More on the VAM Concept

Depending on your league setup, subscribers who use our projections model database, will see the scarcity of quality at these positions (Catchers, Shortstops, Third Baseman, & Pitchers). If you're in a league that sets rosters according to position (and most are), then you have to value the incremental gain by stars (like Ivan Rodriguez C, Alex Rodriguez 3B, Miguel Tejada SS, & Johan Santana P) at a higher premium, since the average player at these positions are worth much less in general than a typical player (in fantasy terms), and some of these players are substantially better than the others at their position. For instance the average catchers' production is approximately 35% less (assuming 200 drafted batters) than that of the average fantasy batter.  In simple terms why pay $20 for Johnny Damon (OF) when you can purchase Ivan Rodriguez for $20. Their fantasy statistics will be comparably balanced, but Rodriguez will outperform the typical catcher by 50%, while Damon will only outperform the typical outfielder by 10%. Thus you capture that incremental gain of 40% (in a perfect world)…doing this over and over again you can see how the odds stack up in your favor (Of course we don't live in a perfect world, thus these gains should not be calculated on a 1 to 1 basis, as you'll read ahead in our Positional Probability paragraph).

After the top 80-90 players are selected, you should start considering your positional and specific category needs. As you can see from the chart above, the incremental gain starts to diminish rapidly at this point. The software will guide you through the favorable VAM players based on your league parameters and then its time to "switch off" and consider the straight dollar values/ or rank as the players remaining to be drafted do not show a significant difference in projected valuation.

Positional Probability

Based on our most recent research, we have refined the VAM model to account for the forecasting probability of every position. Since every sound forecasting model begins with the most recent statistical information as it's base, we performed extensive research on the predictability of using prior year stats as a barometer for  future performance.   Based on our FPI (Fantasy Performance Indicator) calculation, the following chart highlights the 3 year predictability of each position:

3 YR Average Predictability
(based soley on Prior Year Stats)
	Following Season Performance within
Position	15%	25%	35%
C	32%	43%	53%
1B	62%	75%	87%
2B	33%	52%	65%
SS	40%	55%	68%
3B	38%	53%	67%
OF	49%	63%	75%
Starting P	30%	41%	47%
Relief P	37%	48%	55%

As you can see, based solely on prior year stats, Catchers and Pitchers offer remarkably poor predictability. Even at 35% of last year's performance only 53% of catchers and 47% of Starting Pitchers are able to attain mediocre results. While over 87% of 1st basemen reached their mark within the 35% range.

The reason for these results are numerous, but they revolve around logical explanations. Consider that the catching position is one of the most physically abusive positions on the diamond and the unnatural motion of throwing a baseball (pitchers) can take it's toll, making players from these positions a calculated risk. On the flip side, 1st basemen are historically the least likely position players to get injured. Likely due to the relative non physical nature of their playing position.

The bottom line is that these variabilities should be captured in any position scarcity model...and they are captured in ours. It's a big reason why a catcher like Ivan Rodriguez is not the #1 player recommended to be drafted.....he likely would be based solely on a position scarcity model due to the lack of production from the position in general.

The Bottom Line: VAM works

Although VAM would work perfectly if we lived in a vacuum and our projected valuations turned out to be 100% prophetic...of course we don't and they're not going to be.....However through the power of mathematical probability, the odds are in your favor using the VAM strategy.

It does however take more to winning a championship than just drafting an optimal set of players, you will still need to stay on top of the game during the season and make good management decisions (we'll help you here as well).

The VAM strategy is incorporated in our Draft Advisor module contained within the projections software. Our player rankings/valuations adjust as players are selected, adhering to the changing dynamics of your draft (changing scarcity value within the player positions). After a player is selected/drafted, the software will display/suggest the best players remaining....which should guide you to a successful outcome.

 Good Luck on Draft Day, we hope to see you throughout the season !