What is variance in DFS?

Variance is the natural fluctuation in outcomes that happens even when your skill (edge) is constant. A DFS player with a real 5% ROI edge will still have losing weeks, losing months, and occasionally losing quarters — that's variance doing its job. Variance is symmetric: it produces winning streaks that feel like skill and losing streaks that feel like failure, but both are just the noise around your true edge.

How do you calculate your real DFS edge?

Real edge = your win rate minus the break-even win rate. Over a 200+ contest sample, aggregate your actual cash rate (for cash games) or GPP cash rate. Subtract the break-even for the contest type (55.5% for 50/50s, 25%+ for typical GPPs). Positive delta = real edge. This calculation only stabilizes after ~200 entries per contest type; smaller samples are dominated by variance, not edge.

How long can a losing streak last at a real edge?

Much longer than most players expect. A player with a 3% edge in GPPs will experience 8-week losing streaks roughly once per year — not because their skill declined, but because tournament payout variance is enormous and 8 weeks is a small sample. The same player with a 10% edge in cash games experiences 2-3 week losing runs several times per season. Neither is evidence of skill regression.

Should you change your DFS strategy after a losing streak?

Not on the basis of the losing streak alone. If your process was sound before the streak started (real edge over the field, disciplined bankroll management, correct construction rules), the streak is variance and your process is correct. Change your strategy only if you can identify a specific process failure the streak revealed, or if the sample is large enough (100+ contests) to have statistical meaning.

How does variance affect bankroll requirements?

Higher variance requires larger bankroll runway to survive expected drawdowns. Cash games at 1.8x payouts have relatively narrow variance — a 20-buy-in bankroll typically survives normal drawdown periods. GPPs at 1000x+ top payouts have enormous variance — sharp GPP players maintain 50-100 buy-in bankrolls for their target contest tier because 30-buy-in drawdowns happen even at positive edge. See our bankroll piece for the sizing formulas.

DFS Variance vs. Edge: Why Losing Streaks Don't Mean You're Bad

By DFS Degen TeamPublished August 19, 20269 min read

Variance kills more DFS players than skill deficit does. Every profitable DFS player has losing weeks, losing months, occasionally losing quarters — that's the natural noise around a real edge. Players who can't distinguish variance from skill regression tilt into worse construction, chase losses, blow up bankrolls, and quit before their real edge could have compounded. This piece walks through how to compute your real edge, what losing streaks look like at various edge levels, and why "chasing" is the fastest way to convert edge into ruin.

DFS is one of the highest-variance skill games available to a retail player. NBA cash-game player-projection MAE of 5-6 fantasy points on a 50-point mean means the underlying distribution is wide, and lineup outcomes compound that width across 8-9 rostered players. A skilled player with genuine 5% edge routinely loses 4-6 weeks in a row without any process failure. Understanding variance is what keeps the process intact during those runs.

Edge, variance, and how they interact

Your edge is the long-run rate at which your process beats the field. Measured as win rate minus break-even for the contest type. A cash-game player with a real 58% cash rate has 2.5% edge (58% − 55.5% break-even).

Your variance is the natural range of outcomes around that edge. Even at 2.5% edge, individual weeks range from −30% ROI (bad week) to +50% ROI (great week). Over 200 contests, the variance smooths out and true edge emerges. Over 10 contests, the variance dominates the signal.

Weeks tell you nothing about your edge. Months tell you a little. 200+ contests tell you the truth. Every decision below that sample size should be process-based, not results-based.

What losing streaks look like at real edges

Concrete examples from the underlying binomial math:

Cash-game player, 5% edge (60% cash rate)

Probability of losing 3 in a row: 6.4%. Happens every 6-8 weeks on average.
Probability of losing 5 in a row: 1.0%. Happens every 20 weeks (roughly once per season).
Probability of losing 10 in a row: 0.01%. Should not happen in a season; if it does, look for a process failure.

GPP player, 5% edge (25% cash rate vs 20% break-even)

Probability of cashing zero of 4 GPPs: 32%. Happens more than once a week for active players.
Probability of cashing zero of 10 GPPs: 5.6%. Happens ~4 times per season.
Probability of cashing zero of 20 GPPs: 0.3%. Roughly once per two seasons at 5% edge.

The lesson: GPP variance is much wider than cash variance because GPPs cash less often per contest. A GPP-only player experiences apparent losing streaks constantly even at solid edge. The player who "can't buy a cash" in GPPs might be running at 5% real edge and hitting statistically expected variance.

The "chasing losses" trap

Losing streaks trigger a specific psychological failure: increasing bet size to make back the losses faster. This is the classic gambling fallacy — every subsequent bet is independent of the losing streak; the streak has no predictive power over the next outcome.

The math of chasing: if you double your buy-in after a 4-week losing streak, you now need TWICE the win rate to recover — but your win rate hasn't changed. All you've done is amplify variance in both directions while eating into a bankroll that already can't absorb another drawdown. Chasing is how skilled DFS players go broke.

The discipline: bet-size increases only after real bankroll growth (20%+ growth over a 50+ contest sample), not after any subjective feeling of "I'm due for a win." See our bankroll management piece for the promotion criteria.

Computing your own edge

Log every contest you enter for 60+ days. Track: date, sport, contest type, buy-in, entries, payout.
After 200+ contests in a single sport/contest-type combination, compute win rate.
Subtract break-even for that contest type. Positive delta = real edge; negative = you're losing money to variance around a below-water baseline.
Compute the confidence interval on your edge estimate. 200 contests at 58% observed rate gives a 95% CI of roughly 51-65%. Your true edge could be anywhere in that range — the observed number is noisy.
After 500+ contests, the CI narrows enough to make real decisions. Below 500, treat the estimate as directional.

What a losing streak IS worth investigating

Not every losing streak is pure variance. Legitimate process-review triggers:

Your projections systematically miss on one sub-population (e.g., NFL WRs consistently underperform vs. projection). Backtest to check — this is a real model issue, not variance.
Your construction rules broke. Have you started stacking cash lineups, chasing ceiling in cash, or ignoring exposure caps in GPPs? Rule drift is a process failure, not variance.
Field composition changed. If the sport-specific field skill spiked (new pro-heavy contest tier) your edge over that field naturally shrank. Structural change vs. variance.
Rake structure changed. An operator raised rake on your primary contest tier. Same skill, worse per-contest math.

If none of these apply, the losing streak is variance. Don't change the process — the process is what will recover you.

Variance-aware bankroll math

Higher-variance contests require larger bankroll runway:

Cash games: 20-25 buy-ins at your target tier. Cash variance is narrow enough that this survives normal drawdown.
Small-field GPPs (3-max, 20-max): 30-40 buy-ins. More variance than cash but bounded field size caps the extreme drawdowns.
Large-field GPPs (150-max, Millionaire Maker): 50-100 buy-ins. Variance is enormous because payouts are top-heavy and cashes are rare. 30-buy-in drawdowns happen even at real edge; the bankroll has to absorb them.

DFS ROI math — the edge calculation this piece builds on
DFS bankroll management — the variance-aware sizing framework
How to backtest DFS lineups — the discipline that distinguishes process failure from variance
Cash game strategy — narrower-variance surface where edge stabilizes fastest
How many DFS lineups — volume + variance math for GPPs