Glossary
These stats are specific to each sport (e.g. NHL shots on goal, MLB hits, NFL sacks, NBA rebounds) and represent the strategy and tactics of teams attempting to produce performance metrics (i.e. scoring points).
Unless otherwise stated, all stats are calculated as the average (mean) across all games in the currently applied filter.
See Power Ranking for details on how these stats are formatted.
See Offensive vs. Defensive Stats for more info about how stats are framed and ranked.
This table shows the previous 20 matches for each team as well as the previous 20 head-to-head encounters between the two teams. The 'team' columns show what the winning probability was when the match started, with the favored team bolded. The score columns use a color gradient based on the score differential. This can quickly be used to determine a teams form, where upsets happen (i.e. when the highlighted score is on the opposite side of the bolded team name), and how close recent games were.
A color gradiant is used for the scores based on the margin of victory (aka diff between both scores).
The middle column shows '@' to indicate when the left-column team was on the road, thus making the right-column the home team. When there is no '@' symbol shown, then the right-column team was the away team.
The preview page is the core feature of DecodedSports.com. It was the driving idea behind creating the site after identifying a market gap in sports analytics sites that default to simple metrics without context for their previews.
We show a pre-match preview, meaning all the stats shown are as they stats existed the minute before the match started. We take painstaking steps to ensure no future information is leaked into these previews
When viewing historical matches, these previews are also as they existed at the minute prior to the match. For these historical previews, the only 'future information' we show is the final score of the game to help develop your understanding of how the previews translate to outcomes.
Anonymized Previews:
When using our training feature, we show these same historical previews via an anonymized filter to ensure you are not biased by team names or dates when deciding which picks to make. The match history could technically reveal the true identies of teams and the match date if you were to go through the trouble to reverse engineer specific historical scores, but what's the point of training if you cheat! Learn more about training here.
All stats can be viewed from a defensive or offensive lens. Something that is good for the defense, is bad for the offense, and vice versa. Often times, this concept is most commonly represented as 'For' and 'Against', e.g. 'rushing yards for' and 'rushing yards against', or 'points for' and 'points against'.
The concept of 'For' and 'Against' sometimes gets confusing without knowing what lens the stat is being viewed from, i.e. 'For Offense' vs. 'For Defense'. For many stats this is obvious, like 'NFL Third Down Conversion' is most obviously interpreted as an offensive stat, but some stats are less obvious depending on how they are phrased — we frame NFL Sacks and Interceptions from the defensive lens (better when higher). Some stats could apply to both offense and defense like 'NFL Penalties', and 'NHL Shifts', in these cases we convert the stat to be from the lens of which side most commonly generates the stat, e.g. NFL penalties could occur for both sides, but are most commonly committed by the defense. When in doubt and when a stat could be viewed both ways we frame it from the offensive lens by default.
Somes stats are better when they are high (e.g. offensive rebounds), some are better when they are low (e.g. offensive turnovers). If a stat definition says `OffensiveHigher values are better.', then that means when it is viewed from a defensive lens that lower values will be better, and vice versa. This is what determines ordering in Metric Power Ranking.
Most filters on the matchup preview page allow you to view both the offensive and defensive side of a stat. This glossary is your quick reference to see the default lens a stat is framed from and if lower or higher values are better from that lens.
Some places where we show team abbreviations or statistics on this website, we also show the percentile ranking in parenthesis next to it. In these cases, we deemed that using percentile rankings instead of ordinal power rankings (e.g. #1 team, #12 team) was better to maintain a consistent representation since leagues may differ in size season to season or between sports. e.g. #20 may be considered worst in the 1980s, but mid-pack by 2010 after a dozen more teams have been added. In both cases, a percentile ranking of '(0)' would represent worst without confusion.
A percentile ranking of 'ABC (73)' should be read as: 'Team ABC is better than 73% of teams'.
A percentile ranking of 'NHL Shots: (45) 25.12' should be read as: 'Team ABC averages 25.12 shots per game which is better than 45% of teams'.
If a percentile ranking is '0', that means the team is the worst. Conversely, if a percentile ranking is '100', the team is the best.
Percentile rankings are shown for a team's standings in the league (like on the home page) as well as for their rank for specific stats (like on the matchup preview page).
Rankings add important context that is not always conveyed by probabilities or metrics. For example, a #1 team may have a 60% chance of beating a #5 team. Similarly a #20 team may also have a 60% chance of beating a #24 team. The lower ranked teams may not be as consistent as the higher ranked teams though, thus conveying different assumptions about the probabilities and metric averages (i.e. the shape of the distribution, standard deviation, etc).
This table shows the roster of both teams combined. We combine them so they can be directly compared to easily see which players outperform the others on average. This adds more context to understanding lineups compared to traditional methods of representing rosters in their individual team tables.
We keep this table as simple as possible by only showing a few key metrics that are universal to all sports. While we could show all the sport specific stats (goals, hits, blocks, rebounds, etc) we believe that it is noise that could distract from the context. Don't worry, we are rolling out other tools on the site to show the nitty gritty.
The metrics we show are:
- Fantasy Points Per Game (FPPG): This number follows the formula of DraftKings.com as close as possible where our stats support. While Daily Fantasy scoring formulas are not the end-all-be-all, they are a pretty decent method of aggregating all information about a player into a single number that reprents their performance contribution. The better this formula, the less need there is to analyze sport specific stats for the matchup preview.
- Standard Deviation (Stdev) of FPPG: This number represents the variability in FPPG. Averages only tell a very simple story. Standard Deviation completes this story by representing potential upside and downside to a player's contribution, assuming a normal distrution of course (big assumption).
- Play Time Per Game (TPG): How much time a player is in the game, not on the bench. This means different things for different sports. Hockey is 'time on ice', basketball is 'minutes played', but baseball is 'plate appearances' and 'pitches' since there is not a direct time equivalent. Likewise for American football, this is 'snap count'. This number is important since it reveals who star players are that the team relies on to perform. It also reveals where FPPG could be artifficially high if a player had a stellar few games, but doesn't get a lot of play time overall.
Power Rankings are similar to normal league standings but leverage a more sophisticated statistical method to calculate the rankings. These are represented as ordinal rankings (e.g. #3, #12) where lower ranks are better (e.g. #1 is always best).
Traditional standings use a variety of methods such as team winning percentage or points for win/loss. These traditional methods are easy to calculate and understand by humans, but ignore a lot of nuance that can dramatically change our understanding of which teams are better than others.
For example, if the worst team beats the best team in a traditional system, the win is credited with only 1-point. Perhaps a better ranking system would grant the worst team 4 points for such a feat and deduct 4 points from the best team.
See Scoring Distributions and Scoring Performance.
Distributions represent the statistical frequency at which different scoring scenarios occur based on the applied filter.
All probabilities should be read as 'X% of historical values are ≤ the slider value', e.g. if the slider shows '80% ≤ 21', that means that 80% of historical scores for that team have been less than or equal to 21. This is also known as the Cumulative Distribution Function (CDF).
We show four different scoring distributions:
- Away Team Score: See Score.
- Home Team Score: See Score.
- Total Score: See Score Total. We merge the individual Team Score distributions of both teams together (via addition, home_team + away_team) to produce a single distribution. This value represents the Over/Under lines shown by sports books.
- Score Spread: See Score Diff. To compute this value, we again merge the individual distributions (via subtraction, home_team - away_team) to produce a combined distribution.
We try five different types of statistical distributions to determine which is the best fit for representing scores
- Normal (Gaussian)
- Logistic
- Student's T
- Chi Squared
- Weibull Min
For our calculation of Total Score and Score Spread, not all distributions can be combined so easily. The properties of a normal gausian distribution allow this (see Summing Two Normal Distributions) and a Logistic distribution can be similarly aproximated. For all other distributions, however, we must resort to simulation, i.e. we randomly generate a few thousand scoring combinations from each distribution, then determine which distribution fits the result of the simulation best.
You can hover each slider to see a popup with information about the underlying distribution.
Performance focuses on scoring and winning. These scoring based stats are standard across all sports and are agnostic to tactical level box score style stats that are specific to each sport. Performance represents the totality of a team's achievements and their ability to get results.
See Power Ranking for details on how these stats are formatted.
See Offensive vs. Defensive Stats for more info about how stats are framed and ranked.
Metrics can be sliced and diced in a variety of ways. We slice them by 'segment', which is our way of generically referring to the different names sports give time period divisions (e.g. Quarters, Innings). Some segments are common to all sports (e.g. Full-time, Regular Time).
Segment stats and scores are always based on how they stand at the end of a segment, not during a segment. This is particularly important for calculating scoring based stats
All of our stats can be filtered by segment.
- All:
- (FT) Full Time: includes overtime
- (RT) Regular Time: does not include overtime
- (FH) First Half: aka Half-Time, some sports have an explicit half-time (e.g. Soccer, Baseball, Football, Basketball), others do not (e.g. Hockey, Baseball). For those that do not, we use this formula, First_Half = RoundUp(Total_Segments / 2). For baseball this means we use the ending of the 5th Inning as half-time. For hockey, it is the end of the 2nd Period.
- (SH) Second Half: Begins with the segment after the final segment of the first half, e.g. for baseball this is the 6th Inning, for hockey this is the 3rd period.
We only use three-letter abbreviations for teams.
Unfortunately, there is no industry standard for abbreviations or formal lists published by leagues. Each league and media outlet often maintain their own lists, meaning sometimes team abbreviations match and sometimes they do not, e.g., NHL's Montreal Canadiens is 'MON' on CBS Sports, but 'MTL' on ESPN.
The art of abbreviating team names is further complicated when working with historical data. For instance, there are three historical MLB teams to carry the 'Baltimore Orioles' name, each at different time periods (1882-1899, 1901-1902, 1954-present) and treated as different franchise legacies, meaning they each need their own abbreviation. We assign the abbreviations BOR, ORI, and BAL respectively, prioritizing 'BAL' for the most current version of the team since that abbreviation is most recognized by media today.
Algorithmically, we can assign abbreviations using templates like below, assigning more appealing formats to the most recent teams and using less appealing formats near bottom of list for older teams. Fun fact: it takes a list of ~20 such templates to guarantee uniqueness for all historical teams.
e.g., "West Meadowland Unicorns" would be abbreviated as "WES" or "WMU" using the first two formats.
- ABC- ---- ----
- A---- B--- C---
- A--- BC-- ----
- A--- ---- BC--
- AB-- C--- ----
- AB-- ---- C---
- etc
This table shows the Power Rankings for each team and the probability (aka odds) of the home team winning the game.
We do not show the probability for the away team to maintain a cleaner UI and since it is the inverse of the home team probability, 100 - home_team_probability
Percentage of games where the team has led atleast one segment or won the overall game. Note that this measures the score at the end of a segment, not during a segment. If a team was losing at some point during a segment, but was winning at the end of the segment, then they are credited with a lead.
If a team is behind at the end of every segment, but wins the overall game (Full Time), then they get credit for leading at any point.
If a team wins an individual segment but was not leading overall, then they are not credited with the lead