Desirability Index in Blackjack
The Desirability Index (DI) compares expected return with volatility for one completely specified blackjack approach. It is a risk-adjusted comparison metric associated with Don Schlesinger’s Blackjack Attack. DI does not supply a counting system, create an advantage or predict a session.
On this page
Where the Desirability Index comes from
Schlesinger developed DI to compare card-counting games whose return and risk could not be judged fairly from house edge alone. Later standardized comparisons use SCORE, which QFIT describes as DI squared under its defined assumptions.
The important word is compare. A DI number is meaningful only when the simulation inputs and units are defined: rules, penetration, counting system, indices, betting ramp, spread, playing or back-counting policy, players and rounds. Results obtained under different assumptions are not automatically comparable.
Definition and consistent units
One common expression uses expected win and standard deviation over the same 100-round horizon:
DI = 100 × expected win per 100 rounds ÷ standard deviation per 100 rounds.
An equivalent per-round expression is DI = 1,000 × expected win per round ÷ standard deviation per round, under the usual variance-rate convention: expected value scales linearly with rounds, while standard deviation scales with their square root. This conversion assumes the model supports that scaling; do not apply it blindly to correlated outcomes or differently defined simulation outputs. Mixing expected win per round with standard deviation per 100 rounds produces the wrong number.
Example: suppose a simulation reports an expected win of 50 units and a standard deviation of 500 units per 100 rounds. DI = 100 × 50 ÷ 500 = 10. If both quantities are expressed in dollars rather than units using the same conversion, the ratio remains 10. These are illustrative inputs, not a forecast or a recommended game.
QFIT documents the standardized relationship SCORE = DI². In the example that would be 100, but calling it a formal SCORE requires the standard SCORE assumptions and optimized betting conditions. Squaring any informal ratio does not by itself make the result comparable.
What DI can and cannot compare
DI can help compare two simulated opportunities when each has a positive expectation and both use consistent assumptions. A higher expected win increases DI; greater standard deviation reduces it. Penetration, rules, bet spread and play strategy can change both quantities.
DI does not answer whether a person can execute the count, obtain the modeled conditions or tolerate the financial risk. It also does not compare negative-expectation recreational games into profitable ones. A positive model can fail in practice if the rules, penetration, spread, index set or accuracy differ from the simulation.
Use the full simulation output alongside DI: expected value, average and maximum stake, standard deviation, risk of ruin, N0, number of rounds and all assumptions. A single ranking number necessarily hides detail.
How counting and game selection fit
Card counting estimates changes in a persistent shoe. To establish a modeled advantage, a complete analysis still needs rule-matched basic strategy, a validated count and deviations, usable penetration, a tested betting ramp and a risk model. No combination is “surefire,” and finite results can remain negative despite positive expectation.
DI belongs after those inputs are specified. It helps compare complete models; it is not an advanced move layered onto a weak or undefined method. Better knowledge can improve decisions and reduce avoidable errors, but it does not guarantee higher profits or consistent wins.
Ordinary RNG blackjack that generates an independent result each round usually lacks the persistent composition required by traditional counting. Some live-dealer games use physical shoes, but that alone does not establish adequate penetration or a permitted and practical spread. Examine the actual game.
Using DI responsibly
Record the source and version of every result. Do not compare a play-all six-deck simulation with a back-counted double-deck result as if the DI labels make their assumptions identical. If you change the bankroll convention, bet constraints or strategy in a way that changes the modeled bets or play, rerun the analysis. Merely converting every stake and result from units to dollars does not change DI.
DI is designed for positive-expectation advantage-play comparisons, not for deciding how much recreational loss is affordable. A separate budget limit can cap exposure but does not improve DI or expected value. See the risk-of-ruin guide and the SCORE explanation.
Worked comparison: same return, different volatility
Assume two otherwise comparable simulations both report an expected win of 40 units per 100 rounds. Model A reports a standard deviation of 400 units; Model B reports 500 units.
| Model | Expected win / 100 rounds | SD / 100 rounds | DI |
|---|---|---|---|
| A | 40 units | 400 units | 100 × 40 ÷ 400 = 10 |
| B | 40 units | 500 units | 100 × 40 ÷ 500 = 8 |
DI ranks A higher because the modeled return is achieved with less volatility. It does not say A wins every 10 rounds or B wins eight times. Before comparing the numbers, confirm consistent units and round definitions, and document the rules, bankroll convention, spread and constraints for each model. Rules or penetration may differ when those are the conditions being compared; do not confuse an intended difference with an undocumented change in betting policy. Under the standardized relationship documented by QFIT, the corresponding SCORE values would be 100 and 64 only when all SCORE assumptions are satisfied.
What the comparison means for time and bankroll
Increasing every wager by the same factor does not improve this ratio. With the example expected win of 50 units and standard deviation of 500 units per 100 rounds, doubling all modeled stakes changes them to 100 and 1,000 units. DI remains 100 × 100 ÷ 1,000 = 10. The dollar swings nevertheless double. A fixed bankroll now has less room to absorb them, and actual table limits or minimums may prevent proportional scaling.
QFIT also describes N0 as the modeled number of rounds at which accumulated expectation equals accumulated standard deviation for a fixed spread. Using the same example and square-root scaling, after 10,000 rounds the expectation is 50 × 100 = 5,000 units and the standard deviation is 500 × √100 = 5,000 units. Equality is a fluctuation benchmark, not a deadline by which losses must disappear or a guarantee of being ahead.
If those are played rounds in a play-all model, 10,000 rounds at an assumed 50 rounds per hour require 200 hours; at 100 rounds per hour they require 100 hours. Breaks, travel and unavailable tables add real time. For back-counting, distinguish observed rounds, played rounds and elapsed hours according to the simulation’s definitions. A strong per-round comparison can still correspond to a poor practical opportunity when little eligible play is available.
Frequently asked questions
Does DI replace basic strategy or counting?
No. DI ranks the modeled return relative to its standard deviation after the strategy, count, rules, penetration and betting approach have been specified.
Which inputs can change DI?
Anything that changes expected win or volatility can matter, including rules, penetration, count and indices, bet spread, ramp, table occupancy and whether the player observes or plays negative counts.
Is DI useful for online blackjack?
Only if the product supplies a persistent, analyzable card sequence and the remaining assumptions can be modeled. Per-round RNG games are not traditional shoe-counting opportunities; a video of physical cards alone is not enough.
