FREE MLB Player Forecasts,
About the Guru's MLB Player
Projections, Rankings - The spreadsheets below have
6 tabs: Definitions (descriptions of stats), Demo (batter
details), Batting forecasts, Pitching forecasts, and Cheat
Sheets for batters and pitchers. The names of the
forecasted stats have an "m" in from of them. So, the HR
forecast is mHR, etc. Players with no MLB experience do
not have forecasts.
SPECIAL NOTE: The 2020 season data
used for 2021 forecasts are based on 60 regular season
games plus 102 projected game stats. These were the
numbers used: mlb2020.xls
AI detected new pitcher usage patterns and classified a group of what
used to be starting pitchers as LP. There were two cases of this.
First, consider the simple case of Buehler averaging less than 5
innings pitched per start. The machine hedges its bets that he could
continue that pattern as it upholds SP ratings for only the pitchers it
is most sure about. Second, consider the more drastic change of
managers using "openers" who go one or two innings. The AI chose not
classify them as SP either. Interesting times. Note: on
11/28/2020 pitchers were updated with new usage assumptions to account
for unusal managerial decisions during the covid shortened season.
For daily injury updates and
when players are due back: MLB
- Scoresheet Drafting
Spreadsheet (Includes SSL ID's, range, platoon
adj, extra forecasts)
2021 Fantasy Baseball Drafting Tool
(Posted January, 29 2021)
Advice from a human - Don't draft AP
(Avoid) Pitchers even if the machine likes
one. Probably shouldn't draft BP (Bad)
Pitchers either. Those designations are
very telling. JP (Junk) Pitchers are
risky, but you could get lucky with an
undiscovered rookie or star returning from
Forecasts are based on data from the last
three years, where recent years count more. Some of last
years stats are included as a reference. Some stats are
our own invention (described below). And, for pitchers, 2
and 3 year forecasts for key stats are included, but
take these with a grain of salt.
For ranking players, the mGURU stat
is the key. Alternatively for batters, use mGURUwt,
which weighs by positional scarcity. mGURU is
designed to weight the players rate of production and
expected playing time to arrive at a best rating. Since
forecasting playing time is difficult (although we are
proud of the strength of this forecast), mGURU is weighted
more heavily on rate than tradition linear weights. This
strategy is useful for the goal of ranking player
forecasts for simulation leagues.
mGURU is calibrated to value batters and
pitchers roughly equally, possibly giving a slight edge to
pitchers. It does not consider fielding. For batters, SB's
are not as valuable in real life (or in mGURU) as they are
in rotisserie, so draft accordingly.
mROTO is designed to provide values
for 5x5 rotisserie scoring as follows: Batting is Avg, HR,
SB, RBI, R. Pitching is ERA, W, SV, WHIP, Strike outs.
mROTOwt weights values by positional scarcity. Values are
meant to be relative to each other and provide directional
guidance. Values resemble mixed leagues with a $260 budget
for 12 teams.
NL players are listed at the top, AL at
the bottom. If you are playing in a mixed league, for
batters: sort by Play (descending), POS (descending),
mGuru (descending); for pitchers sort by Play
(descending), POS (descending), mGURU (descending). In the
case of pitchers, notice that positions were newly defined
to distinguish between them (see below).
Players can also be sorted by rankings. Rank
is overall for batters.lrank is the league rank
among either batters or pitchers. prank is the
positional rank within league for either batters or
pitchers. An overall rank across all players was not made,
but it can be infered using mGURU.
If you prefer not to consider playing time
in the rankings and want to reply soly on rate of
production, we have that too. For batters use wOBA,
which is based on research by Tom
Tango. For Pitchers use mDERA, which is the
ERA forecast. It relies heavily on Voros McCracken's DIPS
concept. ERAi is the indexed to league ERA.
No minor league data was considered, so be
advised that forecasts for rookies tend to be weighted
more heavily towards league averages for players of their
positions and expected playing time than established
players. Another stat Tom Tango developed called
Reliability will help you gauge this. Best of all is to
look at the 3 similar players players the machine
Looking for sleepers?
Value is S=Sleeper,
R=Regular, and O=Overpriced compared to what
others might expect based on last season. If you
want to know how big a sleeper or how overpriced,
use gurud, which is the difference between
forecasted performance and last year for those
keys stats. The higher the number the bigger
value, the lower the number the more overpriced.
Pitchers: DIF is mDERA-ERA
(forecasted era minus last years era). Negative
dif is a value pick, positive is premium. DIF_IP3
is the change in projected IP 3 years from now
minus last year.
Before we get into the numbers, consider the
machine learning text. It tells us which players are the
most similar using an original approach and clusters them
and gives us a brief description of their status.
ORIGINAL BATTING STATS
The xGURU and xwOBA for
batters shows expected batting value for someone at that
position and playing time. So if you want to consider
valuing SS over OF, then see how much higher that persons
mGURU is above xGURU. mGURUwt was created to do
this weighting for you.
mStarter is predicted playing time,
where Y is Starter; N is Not full time Starter or Inj risk
or Performance Risk; and B is a Bench player with fringe
mlb-level quality forecasted.
mVORP is Value (runs contributed)
Over Replacement Player. It is similar to WAR, but uses a
different bar for replacement value. For batters this is
the wOBA of non-starters at each position. The formula is
(mwOBA-xwOBA)*mTAP. This is different from mGURU, which
further weights the results against players of similar
playing time. mGURU is also weighted more towards rate of
performance, so times at plate plays less of a factor
(although it still matters).
ORIGINAL PITCHING STATS
Creation of pitching positions:
SP=Starters / LP=Long Relief (Swing) / JP=Junk (or Junior)
Starters / EP=Elite Relief Pitchers / DP=Decent RP /
BP=Bad RP/ AP=Avoid RP. These assignments are critical for
drafting. Seriously, do not draft AP or JP pitchers, no
matter what the machine thinks. And, LP might also stand
for "Last to Predict" because I don't know what to do with
these guys (neither do most others, but at least Iadmit
it). I don't have supporting minor league data and LPs are
players in flux. Look for bargains and young talent there.
mSTUFF (Do NOT confuse this with
PECOTA's stuff). Having big STUFF means excellent power
and control. Its our way of measuring upside, especially
for young pitchers. Find hidden future stars! The formula
is: 6*mDOM -2.67*mDERA -3*mCTL -5*mHR9; if mSTUFF<1
then mSTUFF=1; mSTUFF=mSTUFF*mSTUFF/10;
WRIP - Since hrs allowed are more
under a pitchers control than hits allowed, a stat called
wrip was created to look like whip. Whip is
(walks+hits)/ip and Wrip is (walks+hr*10)/ip. I also like
the names because a pitcher giving up lots of hits gets
"whipped", while one giving up lots of HRs gets "wripped."
mVORP is Value (runs allowed) Over
Replacement Player at same position. It is similar to WAR,
but uses a different bar for replacement value. It is
(mDERA-xDERA)*mIP. xDERA ranges from 6 for someone who
only starts to 5 for somone never starts. Like WAR, which
is based on FIP instead of ERA, this is based on mDERA,
which is much more like FIP than ERA.
traj indicates the trajectory the
pitcher is heading over the next 3 years.
Several more are borrowed from others, such
as: mCMD (9*bb/ip), mDOM (9*k/ip), mCTL (k/bb), and HR9