Backtesting Issues in ONE (Part 3)

Last time, I delved deeper into some crazy-looking data I have seen in OptionNet Explorer (ONE) and why it makes little sense. Today I will finish up.

One thing I wonder is whether part of this has to do with inaccurate quotes approaching the close. For example, on 5/7/21 at 3:55 PM EDT, SPX was ~4231. With 7 DTE, the 4275 put (OI = 22) is listed at 46.60 / 54.40. At $7.80, that is wide enough to drive the proverbial truck through! Another trader said:

     > I don’t know what BD you use, but the SPX bid/ask spread on an option
     > 7 DTE with Fidelity is seldom greater than $2—usually between $1 – $2.

This was puzzling because I see wide spreads like this often in ONE. I backed up to 3:30 PM and with SPX ~4236 (OI still 22), the same 4275 put shows 48.40 / 49.00. What an enormous difference! I would trade the latter all day long. With the former, I’d be concerned that slippage might completely offset my edge.

I have heard many times that bid/ask spreads frequently widen after the close rendering EOD quotes unreliable. Could this apply to 3:55 PM? I would actually think this would be smack dab in the busiest, most liquid part of the trading session when bid/ask spreads are narrowest and quotes are most accurate.

EOD liquidity may also pertain to quote freshness. I’m not talking about Sara Lee bread here, but rather how recently the option last traded. An option that has not traded for days (if ever) will likely show a stale quote. Quotes from liquid times of the day are less likely to be stale because higher-volume trading is taking place, but this is a probability: not a guarantee. In the example above, constant OI (22) suggests the option may not have been traded [between 3:30 and 3:55 PM], which then begs my question why the spread would be jacked wide open 25 minutes hence.

I will not backtest with quotes that violate strike arbitrage. I simply cannot believe I would actually see those numbers in real-time despite ONE Customer Support claiming this to be the case [and if I did, then I should immediately arb it regardless of what my current strategy calls for]. Even if I could salvage the exorbitant bid/ask spreads seen on ONE by moving the whole quote up or down to get a valid option chain, it would require spreadsheet manipulation and be a hassle to do.

Open Interest (OI) may be germane to this discussion. OI is 22 in the example given above. This is relatively high compared to what I see on the DITM puts in 2007-2009, which often have slim to no OI. This makes me wonder about a couple things. First, with such low OI am I seeing stale quotes that would change markedly were I to actually place an order? Second, is low OI a real cause of larger slippage? Given ONE data, round-trip slippage for ~5% ITM SPX naked puts in 2007 averages just over $1.30. With slippage factored in, maybe the sticky extrinsic no longer violates strike arbitrage.*

One final consideration regards a methodological tweak I applied in backtesting before Weeklys became available. These particular trade guidelines called for selling 1% ITM with 7 DTE. Before Weeklys, I sold 4-5% ITM on options 4-5 weeks to expiration. The former may be much more liquid than the latter for two reasons. First, near-dated options are usually more liquid than far-dated. Second, expected move is not directly proportional as applied here but rather proportional to the square root of time. Relatively speaking, I was therefore using options even deeper ITM.

I may not be able to backtest with these quirky option prices at least until I have more experience trading similar options to know what usual slippage and bid/ask spreads are like. For now, I should trade small and consider starting my backtest later in the data set when quotes are more reliable. Once I have gained a familiarity with these particular options, I can apply what I know to correct any seemingly distorted quotes from ONE.

* — This would not be the case if bid/ask spread is also similar across a range of strikes.