Gold Digger · An Essay · Screwcap Games
Skin in the Simulation
A conviction you refuse to size is not a forecast. It is a vibe. Why we steer new players straight into a historical scene and make them actually deploy.
Our earlier note, Earned Conviction, argued that a trader should learn the past before betting the present. This one argues something sharper and, frankly, more uncomfortable: that learning does not begin when you read the scene — it begins when you size a position in it. Direction is the cheap half of trading; an undergraduate can be bullish. The expensive half is allocation — how much, with what confidence, at what risk of ruin — and it is precisely the half that spectating cannot teach. We make the case, grounded in Markowitz, Kelly, prospect theory, and the economics of skin in the game, for steering new players immediately into an anonymised historical scenario and requiring them to commit meaningful capital. The discomfort is not a side effect. The discomfort is the curriculum.
§1 The 36-hour delusion
There is a specific and seductive feeling, familiar to anyone who follows markets closely, of knowing exactly what the next day and a half holds. The dollar is soft, the inventory print lands Thursday, the chart has that look — and the next 36 to 60 hours arrange themselves into an obvious story. The feeling is vivid, it is detailed, and it is, with humbling regularity, wrong.
It is wrong for reasons the literature named long ago. Short-horizon forecasting is the natural habitat of overprecision — the flavour of overconfidence in which our confidence intervals are far too narrow for our actual hit-rate (Moore & Healy, 2008). It is sharpened by the planning fallacy and the inside view, in which a richly imagined single path crowds out the boring base rates that would have warned us (Buehler, Griffin & Ross, 1994). And it is anchored by whatever we read most recently, in the way Tversky & Kahneman (1974) made unforgettable. The near-term call feels like knowledge precisely because it is built from the most available, most vivid, least representative information on the table.
We mention this not as detached observers. The clean articulation of this essay arrived the morning after one of us was entirely certain about a two-day move in commodities, deployed accordingly in the only market that bites back, and was relieved of the position and the certainty in roughly equal measure. The tuition was real. This paper is, in part, an attempt to make that tuition available to new players for free — which is the whole business model of a trainer.
§2 Direction is the cheap half
Ask a newcomer what they think gold will do and they will tell you. Ask them how much of the book they would put behind that view and you have asked a genuinely different — and far more revealing — question. The first answer is an opinion. The second is a position, and only the second can be right or wrong in any way that matters.
Modern portfolio theory has been making this point since Markowitz (1952) formalised selection as a trade-off between return and variance rather than a beauty contest of individual picks. The instrument you choose is a small decision; the weight you assign it is the whole game. The sharper and more unsettling formalisation is the Kelly criterion (Kelly, 1956; popularised for markets by Thorp, 2006), which says that the optimal fraction to stake is a precise function of your edge and your odds — and that betting too large relative to your edge guarantees ruin, while betting too small leaves money on the table. Both errors are sizing errors. Neither is visible from the spectator's chair.
Two traders can hold identical views and have opposite careers. The difference is never the view. It is the size.
This is why a trainer that scores only direction trains only the cheap half. A player can be right about gold for two years and bankrupt at the end of it, because they sized a strong conviction like a weak one and a weak conviction like a strong one. The skill we actually care about — the one that separates a hobbyist from an allocator — lives entirely in the gap between the call and the size.
§3 Skin in the game, even fake game
You cannot learn to size by watching. Sizing is a felt skill, and feeling requires stakes — which is why we ask new players, almost rudely early, to deploy 50% or 100% of their (simulated) book.
The behavioural case is overdetermined. Prospect theory established that losses loom larger than equivalent gains by roughly two to one (Kahneman & Tversky, 1979); that asymmetry is not a bug to be designed around but the very nerve we need to put under load, because it is the nerve a real position pulls on. A player who risks nothing feels nothing, and a player who feels nothing learns nothing about the only variable that matters. The point generalises beyond markets: Taleb (2018) made an entire moral philosophy of the observation that judgment without exposure is theatre. Our version is smaller and more practical — but it is the same observation. Conviction is only tested when something is staked against it.
Requiring deployment is also a commitment device in the technical sense (Bryan, Karlan & Nelson, 2010; cf. Thaler & Benartzi, 2004). Left to their own devices, new players hedge, dabble, and nibble — they place a position so small it cannot teach them anything, precisely to avoid the discomfort that is the lesson. Forcing a meaningful deployment removes the escape hatch. It converts a spectator who is "kind of bullish" into a participant who has actually said, with their book, exactly how bullish — and who will therefore actually remember the outcome.
Permit micro-positions and players will use them to launder uncertainty into the appearance of activity. A 2% nibble feels like "playing," but it teaches nothing about size, risk, or conviction — it is spectating with extra steps. The forced 50/100% deployment in the opening scenarios is not a difficulty setting. It is the mechanism that makes the lesson land.
§4 Why a historical scene, and why immediately
Earned Conviction opened the new player in a slow, synthetic tutorial — and that remains the right place to learn the shape of the loop. But the instant the loop is understood, we route the player not into the comfortable live present but straight into an anonymised historical scene, and we hand them capital to commit. The reasoning follows directly from everything above.
A historical scene is the only arena that supplies consequence without ruin. The present offers consequence but, for a beginner, at full and unforgiving tuition. A synthetic scenario offers safety but no real-world texture. An anonymised historical regime — revealed one observation at a time, identity withheld until the verdict, as our companion notes detail — offers both: a genuine outcome that actually happened, a real P&L on simulated stakes, and no possibility of recognising the answer in advance. It is the closest thing to a flight simulator that markets allow, and pilots, notably, do not begin in the synthetic cockpit and then go straight to a storm with passengers aboard. They fly the simulated storm first, controls heavy in their hands.
Steer the newcomer into history with capital on the table, and they orient through their own bankroll — which is the only way orientation ever actually happens.
This is also the gentlest available cure for the 36-hour delusion of §1. A player convinced they can read the next two days is best disabused not by a lecture but by being made to size that conviction inside a regime whose outcome is already settled and merely hidden. The market, even in replay, is an unsentimental teacher. It does not care how vivid the story felt.
§5 Scoring the size, not the guess
If sizing is the skill, then sizing is what we score. A player who is directionally right but chronically mis-sized is not good at this yet, and a scoreboard that congratulated them would be lying.
So calibration, in Gold Digger, is two-dimensional. The first axis is the familiar one from Earned Conviction: did your stated confidence match your hit-rate? The second is the one this paper is about: did your deployment track your edge? Over-committing a coin-flip and under-committing a near-certainty are both failures of allocation, and both are now legible in the record. The Kelly framing gives this teeth: there exists, for any stated edge, a defensible range of sizes, and the player's job is to land inside it more often than chance — to make their book agree with their own confidence.
This is harder than it sounds and more valuable than it looks, because it trains the trait that survives contact with real money. Directional accuracy is fragile; it decays the moment the regime shifts. Disciplined sizing relative to conviction is durable; it is, more or less, what risk management is. A player who leaves Gold Digger having internalised "size to your edge, not to your excitement" has acquired the one habit that keeps real accounts alive.
The whole argument, on one card
| Principle | Why it's there |
|---|---|
| Score size, not just direction | Direction is the cheap half; allocation is where skill and ruin both live. |
| Force real deployment (50/100%) | Stakes create the feeling; the feeling is the only teacher of size. |
| History before present | Consequence without ruin — a settled regime gives a real outcome at no real cost. |
| Anonymise, reveal slowly | You can't recognise the answer, so you must actually size the call. |
| Punish the dabbler | Micro-positions launder uncertainty into fake activity and teach nothing. |
| Calibrate confidence × size | The durable, transferable skill is sizing to edge, not to adrenaline. |
§6 How real allocators actually size
We are not inventing a discipline; we are compressing one. A brief field guide to how professionals turn a view into a weight — every one of them an argument that direction is the easy part.
Fractional Kelly
Few serious practitioners bet full Kelly — the math is optimal in the long run but brutal in the short, so most stake a fraction (half-Kelly, quarter-Kelly) to trade a little growth for a lot less volatility (Thorp, 2006). The lesson for a beginner is the humbling one: even when you have a genuine edge, the right size is usually smaller than your enthusiasm wants.
Volatility targeting & risk parity
Rather than size by conviction alone, many funds size so that each position contributes equal risk, scaling down what is volatile and up what is calm. The mechanics differ from a retail trainer, but the instinct transfers cleanly: a position's size should answer to its risk, not merely to how much you like it. The modern treatment lives in Ang (2014).
Conviction-weighting, done honestly
The best discretionary desks scale into their highest-conviction ideas and keep the marginal ones small — but only after a process forces them to state the conviction before the outcome is known, which is exactly the discipline an anonymised, confidence-scored trainer imposes by construction. Even the machine-learning literature has rediscovered the point: bet sizing, not signal generation, is where much of the realised edge is won or lost (López de Prado, 2016).
§7 What forced deployment can't fake
Honesty, again, because the alternative is a lawsuit and a worse product. Simulated capital is not real capital: it cannot reproduce the specific, sleep-destroying weight of money you actually need, and a player who deploys boldly on a play-money book may freeze on a funded one. We narrow that gap deliberately — meaningful deployment, loss-aversion left intact, calibration that rewards discipline over bravado — but we do not pretend to close it. A strong record inside Gold Digger is evidence of a trained habit, not proof of a live edge.
There is also a dosage question we hold open. Forcing 100% deployment too early can teach recklessness as easily as discipline if it isn't paired with calibration scoring that punishes mis-sizing; the deployment requirement and the sizing score are a matched pair, and shipping one without the other would be its own kind of malpractice. As ever: Gold Digger is a simulated, educational environment, not financial advice and not a brokerage. The aim is to make a player's conviction earned, sized, and honest — and to say plainly where the simulation's authority ends and the market's begins.
§ References
- Bryan, G., Karlan, D., & Nelson, S. (2010). Commitment devices. Annual Review of Economics, 2, 671–698.
- Buehler, R., Griffin, D., & Ross, M. (1994). Exploring the "planning fallacy." Journal of Personality and Social Psychology, 67(3), 366–381.
- Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291.
- Kelly, J. L. (1956). A new interpretation of information rate. Bell System Technical Journal, 35(4), 917–926.
- López de Prado, M. (2016). Building diversified portfolios that outperform out of sample. Journal of Portfolio Management, 42(4), 59–69.
- Markowitz, H. (1952). Portfolio selection. Journal of Finance, 7(1), 77–91.
- Moore, D. A., & Healy, P. J. (2008). The trouble with overconfidence. Psychological Review, 115(2), 502–517.
- Ang, A. (2014). Asset Management: A Systematic Approach to Factor Investing. Oxford University Press.
- Taleb, N. N. (2018). Skin in the Game: Hidden Asymmetries in Daily Life. Random House.
- Thaler, R. H., & Benartzi, S. (2004). Save More Tomorrow™: Using behavioral economics to increase employee saving. Journal of Political Economy, 112(S1), S164–S187.
- Thorp, E. O. (2006). The Kelly criterion in blackjack, sports betting, and the stock market. In Handbook of Asset and Liability Management.
- Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131.