Kelly Criterion in WoW: A Useful Framework (Not a Magic Calculator)

You spot a juicy deal.

A stack of Dracothyst is listed far below the usual price.
You have 1,000,000g liquid.
The buyout costs 200,000g.

The real question is not “Is this profitable?”
The real question is:

How much of my bankroll should I risk on this one idea?

That is bet sizing. And bet sizing is what separates rich goblins from broke goblins.

A quick, honest disclaimer (before we get nerdy)

The Kelly Criterion is a real sizing approach used in finance and gambling. It was designed for situations where the odds are reasonably measurable.

World of Warcraft is not that clean. The Auction House has patches, competition, human behavior, and chaos.

So this is not a perfect “plug numbers and print gold” solution.
You are not going to estimate p with scientific precision.

What Kelly can do for a goblin is still valuable:

It forces discipline. It turns “I feel confident” into “how much can I risk without killing my future?”

Think of this as a framework for position sizing, not a magic calculator.

Step 1: Define what “winning” means (before you touch any math)

In WoW, most people define “win” incorrectly.

Wrong: “It will sell eventually.”
Right: “I can exit at a profit within a deadline.”

So define these three items before you size anything:

Deadline (T)
Example: “I want this gold back within 7 days.”
Profit condition
Example: “This is a win only if I exit at 20% net profit after the AH cut.”
Failure rule
Example: “If I cannot hit that profit by day 7, I exit at market and treat it as a loss for sizing.”

Now “win vs lose” becomes real, not vibes.

Step 2: The Kelly formula (the exact thing you are calculating)

Here is the classic Kelly fraction for a win/lose setup:

f* = (b*p - (1 - p)) / b

What the variables mean in goblin language:

f* = the fraction of your bankroll you should risk on this idea
p = probability your trade is a “win” (profit exit before your deadline)
(1 - p) = probability it is a “loss” (you fail the profit condition by the deadline)
b = your net profit ratio if you win
Example: you invest 200k, your net profit is 180k, then b = 180/200 = 0.9

Two important rules:

If f* ≤ 0, skip the trade. Your edge is not real (or not big enough).
In WoW, your p is always uncertain. That is why we usually use Half Kelly.

Half Kelly:
f_half = f* / 2

Step 3: A quick note on “win vs lose” in WoW (why this is an approximation)

Classic Kelly was built for clean outcomes: you win or you lose your stake.

WoW flips are messier. You can break even, take a small loss, or hold inventory for weeks.

So we force a binary rule to make Kelly usable:

Win: you exit at your defined profit level before your deadline.
Lose: you fail the profit condition by the deadline and you exit anyway, treating the outcome as a loss for sizing purposes.

This is not perfect math. It is practical bankroll discipline.

Step 4: Estimating b correctly (net upside)

Most goblins overestimate b because they ignore real costs.

A cleaner way:

Start with your realistic sell price, not your dream price
Subtract the AH cut (and expected friction)
Divide net profit by your buy cost

Example:

Buy cost: 200,000g
Realistic sell: 400,000g
AH cut 5%: 20,000g
Net sale: 380,000g
Net profit: 180,000g
So b = 180,000 / 200,000 = 0.9

Step 5: Estimating p without pretending

This is where most guides become useless. They say “estimate p” and leave you with feelings.

We anchor p to one practical question:

Can the market absorb my position within my deadline?

The absorbability check (simple and measurable)

Estimate three numbers:

Daily demand: roughly how many units sell per day in your market
Your realistic market share: how many of those sales you can capture
Your position size: how many units you are buying

Then:

Expected units you can sell within T = daily demand x your share x T

If that number is comfortably bigger than your position size, p is higher.
If it is smaller, p is lower.

A simple p bracket (so you do not lie to yourself)

Use this as a starting point, then adjust down for patch risk and heavy competition:

p = 0.70 to 0.85: fast markets, easy exits, low competition
p = 0.50 to 0.70: decent markets, but you must defend your price
p = 0.25 to 0.50: slow markets, oversized positions, high uncertainty

Rule of thumb: if you feel tempted to set p above 0.80, default to Half Kelly or even Quarter Kelly.

Quick TSM tooltip notes (two signals that help you sanity-check p)

TSM will not hand you p, but it can help you avoid delusion:

Sale Rate: a liquidity hint. Higher generally means “this sells reliably,” not “you will sell at your target profit.”
Region Avg Daily Sold: a reality check for absorbability. If your position is bigger than what sells during your deadline, your p should drop hard.

If you only remember one thing: big position + low daily sold = low p.

Example 1: Dracothyst flip, sized with Kelly

Bankroll: 1,000,000g
Position cost: 200,000g
b: 0.9 (from the net calculation above)

You define win:

Deadline: 7 days
Profit condition: 20% net profit minimum
Failure rule: exit at market on day 7

Absorbability and competition suggest this is not guaranteed. You choose p = 0.60.

Now calculate:

f* = (b*p - (1 - p)) / b
f* = (0.9*0.60 - 0.40) / 0.9
f* = (0.54 - 0.40) / 0.9 = 0.155...

So full Kelly says risk about 15.5% of bankroll, around 155,000g.
Half Kelly says risk about 7.75%, around 77,500g.

Verdict: you do not need to buy the entire stack to win the trade.
Sizing is how you survive long enough to keep compounding.

Example 2: Market reset sizing (why bankroll size matters)

Scenario:
You want to reset an enchant market.

Bankroll: 1,000,000g
If you win, your net profit ratio is about b = 0.5
You estimate p = 0.70 based on demand, share, and deadline

f* = (0.5*0.70 - 0.30) / 0.5
f* = (0.35 - 0.30) / 0.5 = 0.10

Kelly says risk 10%, so 100,000g.
If the reset truly requires 200,000g, that is 20% and Kelly is warning you: too big for this bankroll.

That is why partial resets exist. You test the ceiling without risking the account.

The correlation trap (portfolio killer)

Five “good deals” can still be one bad bet.

In WoW, many markets move together:

Profession tuning can crash an entire category
A new season or raid can shift demand overnight and whip supply into the market

So treat highly related items as one bet.

If your herb basket is 10% risk, split that 10% across the whole basket.
Do not run 10% five times.

Diversification is not more items.
Diversification is different reasons you win.

Mini Kelly calculator (simple table, not a spreadsheet)

Inputs:

b = net profit ratio if you win
p = probability of a profit exit before your deadline
Kelly f* = (b*p - (1 - p)) / b
Half Kelly = f* / 2

Example table

Thin edge trade (looks safe, actually not)

b = 0.20
p = 0.75
f* = (0.20*0.75 - 0.25) / 0.20 = (0.15 - 0.25) / 0.20 = -0.50
Verdict: skip

Decent edge trade

b = 0.50
p = 0.70
f* = (0.50*0.70 - 0.30) / 0.50 = 0.10
Half Kelly = 0.05
On 1,000,000g bankroll: 50,000g risk

High edge trade

b = 1.00
p = 0.65
f* = (1.00*0.65 - 0.35) / 1.00 = 0.30
Half Kelly = 0.15
On 1,000,000g bankroll: 150,000g risk

If this table surprises you, good. It means you were probably oversizing before.

Conclusion: Kelly is not about being right, it is about not dying

If you bet too small, you grow too slow.
If you bet too big, you eventually die.

The Goblin Capital approach:

Define win and deadline
Estimate b net, not fantasy
Estimate p using absorbability, not feelings
Use Half Kelly unless you have strong data
Respect correlation

Survive. Compound. Win inevitably.

Search This Blog

Goblin Capital | WoW Gold Making Guides and Economy Insights