How Each Tool Actually Works

Every number on Fantasy Blacksmith comes from published projections and average draft position data, and every step is explained here, including how we tested it and where the data runs out.

Draft Companion: how a player gets a rating

In category leagues every player gets a score in each of the nine categories: points, rebounds, assists, steals, blocks, three-pointers, turnovers, field goal percentage and free throw percentage. A category score says how far above or below the average draftable player that player is, measured in standard deviations, where +1 is clearly better than average, +2 is excellent and -2 is a clear weakness.

Draftable means the players a league of your size will actually take: teams times roster spots, from the top of the average draft position list. In a 10-team league with 13-man rosters that is the top 130 players. Averages and spreads come only from that group, and everyone else is scored against the same yardstick, which is why most players outside it come out negative.

Season totals by default: per-game stats multiplied by projected games, so a player expected to miss games is worth less. A per-game switch is available.
Turnovers are flipped so that fewer turnovers score higher.
Field goal and free throw percentages count volume. They are measured as made shots above what an average shooter would have made on the same attempts, so a perfect free throw line on a handful of attempts counts for little.
The rating is the nine scores added together, and zero is the average draftable player. The yardstick is fixed, so a rating does not change as the draft goes on.

Why a rating can sit far from average draft position

Average draft position is where players actually get drafted across every kind of league, so it blends points and category leagues and reflects reputation as well as production. The rating is a nine-category value measure, so the two will disagree. The biggest gaps are players with one very large weakness, such as high-usage guards who turn the ball over a lot and big men who struggle at the free throw line. All nine categories count equally and a very large strength or weakness counts in full, which is the standard approach for nine-category leagues. We tested capping extreme categories: it helped when drafting with no strategy but lost ground once a strategy was chosen, so we left it out.

Punt strategies

A punt strategy re-weights the nine categories before the scores are added. A punted category is weighted at zero, five categories are boosted (two at double weight and three at 1.75 times), and the rest sit at full or half weight. There are seven strategies: Punt assists and steals, Punt free throw percentage, Punt field goal percentage, Punt turnovers, Punt blocks, Punt three-pointers and Punt steals.

By default the first three picks use the neutral rating, then the strategy that best fits your roster is recommended. It compares the shape of your team's strengths and weaknesses with each strategy's pattern of boosts and punts. You can choose any strategy at any time, including from the first pick. In our simulated drafts committing from pick one was slightly worse than waiting three picks, by about 0.1 category wins per matchup.
A boosted category counts at its full weight up to +1.5 standard deviations. Anything beyond that counts at normal weight, so one extreme stat cannot swamp the rest of a profile.
If your strongest category is 25 percent or more ahead of your second strongest, measured against the average team, its weight is capped at 1.5 times. If your second is 25 percent or more ahead of your third, both are capped.

Catch-up weighting

A strategy cannot know your team has drifted below average in a category it wants. For every category weighted at full weight or more, the weight is multiplied up by how far behind the average team you are, measured in the same standard deviation units: the weight is multiplied by one plus the shortfall divided by 1.5, capped at a 250 percent increase. Being 1.5 standard deviations behind doubles a category's weight and 3 triples it. A punted category is never touched.

Position balance and legal picks

A strategy's best players bunch up at one position. Under Punt assists and steals about 60 percent of the top 65 players are centers, and under Punt blocks most of the top 20 are guards. Position balance works against that, and only applies when roster-position enforcement is on.

Value over replacement: players are grouped into guards, forwards and centers, and each group is marked down by the score of the Nth-best available legal player in that group, where N is about a quarter of the teams in your league. A center is worth less to you if the next center available is nearly as good.
Floors and ceilings: each group has a minimum, the slots it fills, and a maximum. At 13 players the ceilings are 7 guards, 7 forwards and 4 centers. Below its floor a group gets a bonus that grows as your picks run out, and at the ceiling it takes a penalty.
Roster slots are PG, SG, SF, PF, C, G and F with every remaining spot utility, and at most four players whose primary position is center. A player you cannot legally add sinks to the bottom of the list instead of being recommended.

Fantasy points formats

A points league has one number to maximise, so there are no categories to weigh and nothing to punt. Players are scored the same way, in standard deviations from the draftable pool, but on fantasy points, as a season total by default or per game. ESPN head-to-head points weights are points 1, field goals made 2, field goals attempted -1, free throws made 1, free throws attempted -1, three-pointers 1, rebounds 1, assists 2, steals 4, blocks 4 and turnovers -2. Yahoo traditional weights are points 1, rebounds 1.2, assists 1.5, steals 3, blocks 3 and turnovers -1, and Yahoo high score drops the turnover penalty. Position logic is lighter in points mode, because with one scoring number and utility slots that take any position, scarce positions are worth much less.

How we tested it

The Draft Companion's scoring was run through simulated drafts: 10 teams, 13 rounds, snake order, with the other nine teams drafting by average draft position plus random noise. Each draft is scored by expected category wins per matchup against the other nine teams across all nine categories. For points formats the measure is simulated daily starting-lineup fantasy points.

Plain strategy score: about 4.9 of 9 category wins, below average in about 3 of the 7 or 8 categories the strategy was meant to be strong in.
With catch-up weighting: about 5.2. With Light position balance on top, the default: about 5.3, and drafts with four or more centers in the first six picks fall from 26 percent to 0.
Points leagues: average draft position order gives +3.8 percent lineup output against the league average and beats 77 percent of teams. Value ranking with Light position balance gives +6.4 percent and beats 92 percent.
The tests also removed things that did not hold up: a 0.8 penalty on centers' blocks was a plain bias against centers and cost about 0.12 category wins, so it was retired.

A simulation is a model, not a guarantee, but it is far better evidence than a hunch.

Post-Draft Analyzer

A finished roster is compared with an average team built by simulating how a league of your size drafts, then averaging many of those simulated drafts together for a stable baseline.

League Analyzer

Every roster in the league is known, so the standings come straight from real projected stats. Rotisserie scores are real ranks in each category, and projected records come from a simulated round robin.

Trade Machine

Each player gets a value from their fantasy points per game and how often they are expected to play. Uneven trades add a waiver-wire filler player, and category leagues check how many categories each side improves.

Where the data comes from, and its limits

346 of the 475 players use ESPN's 2026-27 fantasy basketball projections, with the top 120 refreshed on 29 September 2026. The other 125 deeper bench players use their actual 2025-26 per-game stats from Basketball-Reference.
182 players have ESPN's real average draft position, through about 180th, and the rest are estimated. A standard league's draftable pool sits entirely within the real ones.
Yahoo values apply Yahoo's scoring weights to the same projected stats; they are not Yahoo's own projections. Position eligibility comes from ESPN.
The ratings cannot know about news after the data was pulled, and projections are estimates, with projected games played one of the biggest sources of error. Schedules, playoff weeks and custom league scoring are not modeled.
Fantasy Blacksmith is an independent project and is not affiliated with the NBA, ESPN, Yahoo or Basketball-Reference.