Bootstrapping

When an exchange asks a panel of traders across the country what a commodity's spot price is today, it gets many different answers. Some are close to the real market price; a few are far off, by mistake or because that trader saw an unusual deal.

Bootstrapping is how the exchange filters out the bad answers to arrive at one fair spot price for the day — used, in turn, to keep futures prices honest and to calculate final settlement.

Spot prices for the day's polling are collected from empanelled polling participants — traders and users of the commodity across different centres. From this set of polled prices, bootstrapping first finds a median price and discards the outliers around it.

An algorithm then runs across the remaining data: it repeatedly selects a random price from the set, checks the standard deviation of that selection against the rest of the polled data, and discards the random price whose standard deviation is higher. This is repeated until a final number is reached — the polled price with the least standard deviation against the full polled set, based on the principle of least error, similar to a least-squares regression. The resulting spot polling price may be close to the arithmetic average, median or mode where the polled data is tightly clustered.

On a given trading day, NCDEX collects turmeric spot prices from twelve empanelled participants across Erode, Nizamabad and other centres. Ten quotes cluster between Rs 13,200 and Rs 13,450 per quintal; two outliers come in at Rs 11,000 and Rs 15,800, from traders quoting unusual, thin deals.

Bootstrapping first strips out the Rs 11,000 and Rs 15,800 outliers around the median. It then repeatedly tests random selections from the remaining ten quotes, discarding whichever selection produces the highest standard deviation against the rest, until it converges on a single figure — say, Rs 13,320 — declared as the day's spot polling price. Futures traders across the country then see this one number, not twelve conflicting quotes, disseminated on the exchange screen.

Chapter 3 (Commodity Futures), section 3.10 (Spot Price Polling and Final Settlement Price of Futures), sets out bootstrapping as the method used to clean polled spot prices before they feed into both daily dissemination and the Final Settlement Price calculation. Expect a question on what bootstrapping is designed to remove (outlier prices) and on the criterion used to pick the final number (least standard deviation against the polled set).

• Bootstrapping removes outliers; it does not simply average all quotes. A plain arithmetic mean would let a single bad quote distort the result — bootstrapping is designed precisely to avoid that.
• The final price is chosen by least standard deviation, not by majority vote among the polling participants.
• Bootstrapping is a spot-price cleaning method, distinct from the separate rule that the Final Settlement Price is the average of polled prices over the last three days up to and including expiry — bootstrapping produces each day's single polled figure; the three-day averaging happens afterwards.