Picking up nickels, and the tail

"Picking up nickels in front of a steamroller" is the standard description of short premium, and it is used so often that it has stopped meaning anything. This lesson puts numbers on it. A two-year track record of a reasonable, defined-risk condor program, and then a month like March 2020, side by side. Then the same month for the trader who left the wings off.

Two good years

An index at 3,300 with implied volatility around 14%, a level that was normal for long stretches of the late 2010s. The 45-day expected move is 3,300 × 0.14 × 0.351 ≈ 162 points, about 5%.

A trader sells a 10-delta iron condor each month: short 3,080 put / long 3,050 put, short 3,470 call / long 3,500 call. Thirty-point wings. Credit about $4.50 per condor ($450), max loss 30 − 4.50 = $25.50 ($2,550). Ten condors a month: $4,500 of credit against $25,500 of defined risk.

Suppose the previous 24 months looked like this, which is entirely plausible for that period:

Twenty-one winners out of 24, roughly $940 a month on $25,500 of risk. Not spectacular, but steady, and the kind of record that makes a trader size up.

One bad month

The condor for the March cycle is opened around February 19, 2020, with the index near its high. Over the following six sessions the index fell about 12%. Between February 19 and March 23 it fell about 34% peak to trough, and the VIX closed above 82 on March 16, its highest close on record at the time. These are the real figures.

For the condor program:

• On the first −3% day the short 3,080 put is not yet breached, but IV has jumped and the put spread is worth about $12. Condor value near $12.50; the 2× stop ($13.50) has not quite triggered. Many traders held.
• On the next −4% day the index gaps through 3,080 at the open. The put spread opens near $22 with IV in the 40s. The stop was passed inside the gap; the fill is at about $22.50, a loss of $1,800 per condor, $18,000 for ten.
• A trader who did not act at all reached max loss, $2,550 per condor, $25,500 for ten, within two more weeks.

Twenty-one winning months paid for, at best, one losing week with $4,500 left over. And that is the defined-risk version. The wings did exactly what wings do: they turned a catastrophe into a bad year.

The same month without wings

Now the trader who found the wings expensive and sold ten naked 3,080 puts for about $6.50 each instead ($650 per contract, $6,500 a month; about 45% more credit than the condor). Initial margin roughly $30,000 for the ten.

The index closed near 2,240 on March 23. A 3,080 put is 840 points in the money: $84,000 per contract, $840,000 for ten, against two years of collecting $6,500 a month, about $156,000 gross. But the trader never got there:

1. On the −12% week, the puts were worth about $2,500 each, a $19,000 loss on ten, and the margin requirement had roughly tripled as the strike came into the money and IV exploded.
2. The margin call arrived with the loss. The account could not meet it. The broker closed the position at the market, at IV levels that had not been seen since 2008, paying the widest bid-ask spreads of the decade.
3. The realized loss was set by the broker's liquidation, not by any stop the trader had written down.

This sequence is not hypothetical in kind. In November 2018 a firm that sold naked natural-gas options for clients lost more than the clients' entire accounts in a few days when the futures spiked; clients were left owing money. Undefined risk means the loss is not limited to what you have.

Short-premium P&L distribution (schematic)

   frequency
      |            ###
      |           #####
      |          #######
      |        ###########
      |  #   ################
      +--+---+---+---+---+---+---+---+---+---+
     -25x  ...              -1x  0  +0.5x +1x
      ^                           ^
      the tail (rare, decides    the nickels (most months)
      the year)

What the numbers say

• The win rate is not the edge. Twenty-one of 24 is a win rate. The two-year P&L divided by the one-week loss is the edge, and it was about 1.2 to 1 for the condor program including the good years.
• Stops do not work through gaps. A 2× stop is a plan for an orderly market. The loss on a gap is set by where the market opens, and short gamma is at its worst exactly then.
• Wings are the only thing that made the March loss a number you could write down in advance. They cost about a third of the credit every month for two years. That was their price for capping the loss at $25,500 instead of $840,000.
• Size for the tail, not the nickel. If a full-width loss on all positions at once would end your trading, the position is too big, regardless of how many months it has worked. Module 7 gives the sizing rules.
• IV rank was high all the way down. The filter that said "sell premium now" in the first week of March was pointing at the steamroller.

Key idea: A good short-premium program earns nickels for years and gives most of them back in a week, and that is with the wings on. Without wings the loss is not bounded by your account. Judge the strategy by the tail, size so the full-width loss is survivable, and never let two years of wins persuade you to remove the protection that makes the third year possible.

Try it: Take your own short-premium positions, or a paper book of ten condors. Compute the loss if every one of them reaches max loss in the same week, and express it as a percentage of your account. Then compute the same for the strangle version with the underlying down 30% and IV tripled. Write both numbers where you will see them before your next entry.

Recap

• A 10-delta condor program can produce 21 winning months out of 24 and still lose most of that in a single week.
• Stops fail through gaps; the loss is set by the opening print and the short-gamma exposure at that moment.
• Wings cost about a third of the credit and are the only reason the March 2020 loss was a fixed, pre-known number.
• Naked short options in a crash produce losses larger than the account plus a margin call at the worst possible prices.
• Judge short premium by the tail; size so a simultaneous full-width loss is survivable.