AI Report Card · FFL-BOSS

Record

5-1

You started

842

Advice would have

842

Advice value

Perfect hindsight

984

Followed

84%

Data

Refresh results

Week by Week

Wk	Result	Win prob	Started	Advised	Perfect	Followed
1	W	47%	124.26	123.26	166.3	9/9
2	L	55%	101.48	108.38	122.88	4/9
3	W	51%	143.68	116.58	159.88	7/9
4	W	51%	132.72	158.82	181.42	8/9
5	W	77%	175.4	174.2	185.0	9/9
6	W	57%	164.36	160.36	168.96	9/9
7	not played	54%	0.0	—	—	7/9

Started is what you actually played, read from ESPN. Advised is the app's recommended lineup scored on the same results. The gap between them is what the advice was worth; the gap to Perfect is luck nobody could have captured.

Where It Went Wrong

efficiency

td variance

Each cause implies a different fix. inactive means stale data, not a bad model. td variance means the model was right and the dice were not. volume means a role change the projections missed — the one worth acting on. efficiency is ordinary performance swing.

Week 1

RB	— 88 yards but no touchdown - production came, points did not	19.81	13.8	-6.0
RB	— 69 yards but no touchdown - production came, points did not	16.14	9.9	-6.2
WR	— normal usage, unusual production	16.63	21.2	+4.6
WR	— normal usage, unusual production	15.03	26.0	+11.0
DST	— normal usage, unusual production	7.4	13.0	+5.6
K	— normal usage, unusual production	8.21	1.0	-7.2

Week 2

QB	— 3 touchdowns - the model was right, the dice were kind	18.87	28.48	+9.6
WR	— normal usage, unusual production	13.7	7.0	-6.7

Week 3

QB	— 4 touchdowns - the model was right, the dice were kind	21.12	31.28	+10.2
TE	— normal usage, unusual production	11.27	3.3	-8.0
DST	— normal usage, unusual production	7.71	2.0	-5.7
K	— normal usage, unusual production	8.13	16.0	+7.9
WR	— normal usage, unusual production	14.41	7.6	-6.8

Week 4

WR	— 2 touchdowns - the model was right, the dice were kind	17.04	30.8	+13.8
WR	— normal usage, unusual production	16.71	25.8	+9.1
TE	— normal usage, unusual production	12.64	9.1	-3.5
DST	— normal usage, unusual production	7.72	12.0	+4.3
WR	— normal usage, unusual production	13.89	23.2	+9.3

Week 5

QB	— 4 touchdowns - the model was right, the dice were kind	19.52	24.6	+5.1
RB	— 2 touchdowns - the model was right, the dice were kind	15.82	22.6	+6.8
WR	— normal usage, unusual production	16.66	10.6	-6.1
TE	— 2 touchdowns - the model was right, the dice were kind	12.29	32.4	+20.1
DST	— normal usage, unusual production	5.46	20.0	+14.5
K	— normal usage, unusual production	8.59	3.0	-5.6
WR	— normal usage, unusual production	13.96	29.1	+15.1

Week 6

RB	— 2 touchdowns - the model was right, the dice were kind	21.43	30.7	+9.3
RB	— normal usage, unusual production	18.4	24.3	+5.9
WR	— 67 yards but no touchdown - production came, points did not	17.11	12.7	-4.4
WR	— 79 yards but no touchdown - production came, points did not	17.02	10.9	-6.1
TE	— normal usage, unusual production	14.07	23.9	+9.8
DST	— normal usage, unusual production	6.65	3.0	-3.6
WR	— normal usage, unusual production	15.86	28.5	+12.6

Win-Probability Calibration

When the app says 65%, do you win about 65% of the time? This is the single number that says whether to trust the simulation at all.

Band	n	Said	Actual	Gap
0–20%	0	no weeks yet
20–40%	0	no weeks yet
40–60%	5	52.2%	80.0%	+28
60–80%	1	77.0%	100.0%	+23
80–100%	0	no weeks yet

A fantasy season is 14 games, so expect noisy bands for a long time. Consistent one-directional gaps are the meaningful signal, not any single row.

Projection Accuracy

Player-weeks

Avg error

5.76

Bias

+1.8

Bias is the half worth watching: random error is the game, but a persistent lean means projections are systematically off and can be corrected. Negative = the app over-projects.

Pos	n	Error	Bias
QB	6	5.51	+2.9
RB	12	3.96	+0.9
WR	15	7.23	+2.8
TE	6	7.65	+2.9
K	6	4.16	-1.5
DST	6	5.67	+2.5

Biggest misses

Isaiah Likely	proj 11.27	3.3
Harrison Mevis	proj 8.21	1.0
Malik Nabers	proj 14.41	7.6
Terry McLaurin	proj 13.7	7.0
James Cook	proj 16.14	9.9