The strategy graveyard

You now know how to find an edge, test it honestly, combine it, size it, and manage its risk. This chapter tells you what happens next to most of the few edges that survive all that: they die. Even a strategy that was genuinely real, properly validated, and carefully built has a limited life, and understanding how edges die is as important as knowing how to find them, because a quant spends as much time burying strategies as discovering them.

Alpha decays

An edge that was real can fade, and watching it happen is sobering.

# An edge that was real can DECAY: as others discover it, the premium erodes. Here
# a factor's true monthly premium fades linearly to zero over ten years. Because a
# single Sharpe estimate over a few years is noisy (Part 1's lesson), we average the
# rolling Sharpe over many simulated runs to see the EXPECTED decay cleanly.
import numpy as np

rng = np.random.default_rng(40)
n = 120                                     # 10 years of monthly returns
premium = np.linspace(0.012, 0.0, n)         # 1.2% per month fading to 0 over the decade
vol = 0.03
window = 36                                  # a 3-year rolling window
paths = 4000

sims = premium + rng.normal(0, vol, (paths, n))
rolling = []
for t in range(window, n + 1):
    seg = sims[:, t - window:t]
    s = seg.mean(axis=1) / seg.std(axis=1, ddof=1) * np.sqrt(12)
    rolling.append(s.mean())                 # average across runs
rolling = np.array(rolling)

print("Expected rolling 3-year Sharpe as the edge decays (averaged over many runs):")
print(f"  early (years 1 to 3):  {rolling[0]:.2f}")
print(f"  middle:                {rolling[len(rolling) // 2]:.2f}")
print(f"  late (years 8 to 10):  {rolling[-1]:.2f}")
print("\nThe edge was strong early and is gone by the end. Alpha decays as an edge")
print("gets crowded, arbitraged away, or overtaken by a change in the market.")
print("A quant's job is as much retiring dead strategies as finding new ones.")

Expected rolling 3-year Sharpe as the edge decays (averaged over many runs):
  early (years 1 to 3):  1.21
  middle:                0.71
  late (years 8 to 10):  0.21

The edge was strong early and is gone by the end. Alpha decays as an edge
gets crowded, arbitraged away, or overtaken by a change in the market.
A quant's job is as much retiring dead strategies as finding new ones.

Averaged over many runs to cut through the noise, the rolling Sharpe of a decaying edge falls from 1.21 in the early years to 0.21 by the end, from a genuine edge to almost nothing. This is alpha decay, and it is the normal fate of edges, not a rare misfortune. The word alpha means return above what the market gives you for free, and alpha is perishable. The reasons it decays are worth naming, because they are the causes of death in the strategy graveyard.

The graveyard: how edges die

Most strategies die of one of five things. The commonest is overfitting that was never caught: the edge was never real, just the luckiest of a search, and live trading finally reveals what the deflated Sharpe would have warned. Then comes crowding: a genuinely real edge is discovered by others, more money piles into it, and the premium is competed away, which is the decay you just saw. The better known a factor becomes, the more crowded and thinner it gets. Third is capacity: your own size grows until your market impact eats the edge, so success itself kills the strategy. Fourth is costs: an edge too thin to survive realistic costs and slippage, especially one that trades often. And fifth is regime change: the market itself changes, through rates, participants, or rules, and a real relationship simply stops holding, the non-stationarity from Part 2 made concrete.

Retiring is the job

Because edges decay, a quant must monitor every live strategy and retire it when its edge is gone, rather than cling to it hoping it returns. This is emotionally hard, because you built it and it once worked, and it is essential. The discipline is the off-switch from the risk chapter: mechanical rules that cut or stop a strategy whose live, out-of-sample performance has decayed below its bar, without waiting for hope to be disproved. Finding edges is the glamorous half of the job. Burying dead ones on schedule is the unglamorous other half, and skipping it is how quants hand back years of gains to a strategy that quietly stopped working.

What to carry forward

Even a real, well-built edge dies, its Sharpe decaying over time, which you saw fall from 1.21 to 0.21 as a premium eroded. The causes fill the strategy graveyard: overfitting never caught, crowding, capacity limits, costs, and regime change. Alpha is perishable, so a quant monitors every live strategy and retires it mechanically when its edge is gone, an unglamorous discipline as important as discovery. Given how many strategies die, and how the odds stack up, the next chapter measures honestly who you are really competing against: the professionals, and the gap between them and you.