Anatomy of a Short Squeeze: GameStop, January 2021, Reconstructed From the Tape

A short sellerborrows shares, sells them, and hopes to buy them back cheaper. The danger is asymmetric: a short’s profit is capped at 100% (the stock can only fall to zero) while the loss is unbounded. GameStop was extraordinary because short interest exceeded its free float — more shares were sold short than actually existed to trade freely, a structural impossibility sustained only by re-lending the same borrowed shares. The tape entering the event was sleepy: GME closed at $2.44 on 2020-10-01 and $4.71 at year-end.

A short squeeze is a forced-buyer cascade. As price rises, shorts face mounting mark-to-market losses and margin calls. To cut risk they must buy shares to cover — but buying pushes the price higher, triggering more margin calls and more covering. With short interest above the float and few shares available to borrow or buy, the demand to cover meets a wall of no supply. Price is the only variable left free to move. The daily returns show the cascade compounding: +51.1% on Jan 22, +92.7% on Jan 26, then +134.8% on Jan 27 — followed by -60.0% on Feb 2 once the forced buying stopped.

A squeeze is not buyers beating sellers. It is sellers who are contractually obliged to become buyers, into a market with nothing to sell them.

The 2021 episode added a second, faster engine: options. Retail bought enormous quantities of out-of-the-money call options. The dealers who sold those calls are short gamma, so to stay hedged they must buy the underlying as it rises — and buy more as the calls move toward the money. This gamma squeeze is a feedback loop bolted onto the short squeeze: rising price forces dealer buying, which raises price, which forces more buying.

import numpy as np
from scipy.stats import norm

def call_delta(S, K, T, r, sigma):
    d1 = (np.log(S/K) + (r + 0.5*sigma**2)*T) / (sigma*np.sqrt(T))
    return norm.cdf(d1)

# As S climbs toward the strike, delta -> 1, so the dealer's
# hedge demand (shares per call) ramps from ~0 to ~100/contract.
for S in [40, 60, 80, 100]:
    print(S, round(100 * call_delta(S, K=60, T=0.05, r=0.0, sigma=1.2), 1))

At a $60 strike, the dealer’s hedge runs from 8.4 shares per contract at $40 to 97.9 at $100 — nearly the whole contract bought on the way up, mechanically, regardless of any view on the company.

With daily prices you can quantify what the tape did. Compute the run-up, the realised volatility blow-out, and — the number a risk manager cares about — the loss to a constant short position held through the event.

import numpy as np, pandas as pd

px = pd.read_csv("cs15_gamestop.csv", index_col=0, parse_dates=True)["GME"]
ret = px.pct_change().dropna()                     # daily GME, Oct 2020 - Mar 2021

run_up  = px.max() / px["2020-12-31"] - 1          # +1745% peak vs year-end
rvol    = np.log(px).diff().rolling(10).std() * np.sqrt(252)   # peaks at +952%

# P&L of a $1 short opened at year-end, marked daily (loss is unbounded)
short_pnl = 1 - px / px["2020-12-31"]              # bottoms at -1745%
print(f"run-up {run_up:.0%}   worst short P&L {short_pnl.min():.0%}")

The squeeze did not stay inside one ticker. XRT — the retail-sector ETF whose own short interest made it a squeeze conduit — jumped +41.8% over year-end to its 2021-01-27 peak, and the VIX closed at 37.21 the same day GME topped: a single small-cap forcing a repricing of index-level risk.

A standard Value-at-Riskmodel, calibrated on GameStop’s own pre-event history, saw little of what was coming. On the Oct–Dec 2020 sample the daily volatility was 8.6% and the historical one-day 99% VaR was -14.1%. Jan 27 delivered +134.8% — a 15.6σ day on that calibration. VaR built on a calm sample is blind to a regime change driven by positioning, not fundamentals. The squeeze risk lived in three numbers a price-only model never sees: short-interest-to-float, days-to-cover, and the options gamma profile.

GameStop is not a story about a meme; it is a story about asymmetry and crowding. Capped upside and unbounded downside make a short an option you are selling — and a crowded short with no float is the most expensive option you can sell. A $1 short held from year-end was marked at -1745% of its proceeds at the peak — a 17× loss on a position whose maximum gain was 1×. The transferable lessons: watch positioning, not just price; respect that liquidity vanishes precisely when you need to exit; and never let a model calibrated on calm tell you a crowded trade is safe.