The House Always Wins #1: Loss Aversion

Scores of retail traders lose money systematically – i.e. strategies replicating exactly their opposite trades, offer guaranteed pay-offs, given a statistically speaking representative time horizon, and sufficient capital to withstand the occasional loss.

The question is: how does this come about?

The answer is a cornerstone of behavioral economics. To understand the mechanics in detail, please refer to a fantastic book by Daniel Kahmemann, the Nobel prize winning economist who inspires many of our posts on the series “The House always wins”, including the explanation regarding loss aversion that follows.

Daniel Kahneman: Thinking fast and slow @RIGB

Imagine you’ve sued someone and taken them to court.

You’re looking good: you have a 95% chance of winning and being awarded $1.000.000 in damages, and 5% chance of not receiving any compensation. Out of the blue comes someone and offers to pay you $ 900.000 regardless of what the verdict is, in exchange for the payout.

What do you choose? Your humble blogger (and 99.9% of respondents when posed this question) takes the $ 900.000 and runs. In taking this choice, we implicitly take out insurance for $ 50.000 (the difference between the expected value of $ 950.000 and the $ 900.000 settlement) rather than face the frightening 5% chance of seeing the million vanish.

Behavioral take away #1:

Humans instinctively place value on certainty. Most of us are intuitively willing to pay an expensive premium to insure ourselves against an uncertainty with comparatively small probability, but devastating effect.

Now look at the situation from the opposite angle: assume your attorney offered the party you sued to drop charges in exchange for a $ 900.000 settlement. Put yourself in their shoes, and answer in 5 seconds or less: would you take the deal?

Let us guess: you probably wouldn’t!

This is the paradox of risk aversion: whoever rejected this would give up arguably the best deal in their lives by risking a 95% certain promise of $ 50.000 (the expected value of $ 950.000 minus the settlement) in exchange for a 5% shot at saving one million. Somehow the human mind cherishes the small shot at a huge payout and eschews a certain, but less juicy outcome (95% chance of making 100.000).

Now look at the exact same situation, but instead of looking at it with the eyes of a private individual of limited means, use those of the Chief Financial Officer of a large corporate with USD 1 billion in cash sitting idle, who regularly faces such lawsuits. What do you think he/she’d do?

Yep – our CFO just took the deal. He/she looked at the situation, weighted probabilities and payouts, and took the “right” course of action.

Behavioral take away #2:

if you’re asked to play this game repeatedly and have plenty of cash to sit out the occasional (5%) adverse outcomes, in the long run humans switch on their “statistical mind” and can statistically be expected to make very good choices on a risk adjusted basis.

The applications of both take-aways to financial markets in general and retail trading in particular can’t be overstated. This is the possibly the main reason why most retail traders struggle to close out losing positions AND similarly, why they rush to close out the profits on their winners. The results are well known: the more they trade, the more they lose.

So remember that speculative trading is a 0 sum game: your win is someone else’s loss, and your loss quite likely is the House’s win. Now look at this picture.



We sincerely hope you’ll think of it the next time you hear “The House always wins” 🙂

PS – if you’re wondering if your track-record suffers from loss aversion, feel free to sign-up and get your D-Score, including a 0-10 Discipline diagnosing you for loss aversion, among other afflictions potentially affecting ANY trader 🙂

Inside Darwinex – Performance calculation, explained

Understanding your Performance

Upon joining, many new movement members reach out stating: “hey guys, my performance statistics are different from Service Alfa (insert here your favourite alternative site to Darwinex) – they’re wrong!”

Whilst occasionally stuff DOES go wrong, more often than not differences owe to the fact that returns at Darwinex are calculated differently from most existing service providers.

Other services calculate performance on changes to account balance, which accounts for realised P&L (closed trades), but ignores unrealised P&L until open trades are closed.

Darwinex performance is measured on changes to liquidation value, which accounts for realised P&L (closed trades) as well as unrealised P&L (i.e. open trades).

This is the most common reason why returns “differ”. The rest of this post explains why we do things this way, and what implications this has.

Other services performance calculation – changes in account balance

Most services track changes to balance, i.e. accounting for

  • Starting account balance
  • Plus minus realised P&L
  • Plus minus cash flows
    • Cash paid in / deducted by your broker (swaps, commissions, etc.)
    • Monies deposited in / withdrawn by yourself

Performance over a given reference period is then driven by the % change in the resulting account balance.

Your Darwinex Performance – changes in mark to market

At Darwinex, all strategies’ Trading Journal reflect daily liquidation values, considering:

  • Starting account balance
  • Plus minus realised P&L: i.e. monies made or lost on closed positions
  • Plus minus cash flows as per above
  • Unrealised P&L: i.e. what-if P&L when closing ALL open positions at the end of every day

I.e. at the end of every day, regardless of whether trades are open or closed, we compute the hypothetical P&L for ALL open positions assuming that they were closed at the then going market price. The resulting account “balance” is called liquidation value for that particular day.

Your performance is calculated on changes to this liquidation value (purists would call this Mark-to-Market value) of equity. Which means, on any day where open positions are held overnight, the inputs into your performance calculation are different from the inputs at Service Alfa. Unsurprisingly, the output is also different!

Unrealised P&L – why do we care about it?

Darwinia rates strategies for DARWIN investor appeal (hence Darwinia): our standpoint is that of an investor replicating, with his own monies, all trading decisions in a strategy.

This makes us care about unrealised P&L as much as we care about account balance: let’s work through an example to illustrate why.

Market movements matter

Trader Ace with USD 10,000 balance opens a 5 lot long EUR/USD trade. For the next 3 days, this will be his only open trade.

This is what happens to the EUR/USD over the timeframe.

  • Trade open at 1.30135 on day 0
  • End of Day 0: EUR/USD is at 1.30115. Ace keeps the trade overnight
  • End of Day 1: EUR/USD is at 1.29250. Ace keeps the trade overnight
  • End of Day 2: EUR/USD is at 1.28815. Ace keeps the trade overnight
  • Day 3: trade is closed at 1.30250

Investor A replicates Ace’s trades with his own monies 1:1, i.e. using the same leverage as Ace. A uses Darwinex and service Alfa to monitor his investment.

This is the performance calculation investor A gets at Darwinex and Service Alfa [1]:

Snapshot EUR/USD TradeSlide Service Alfa
Liq. value  Daily Perf. Balance Daily Perf.
Trade open 1.30135 10,000.00$ N/A 10,000.00$ N/A
End of day 0 1.30115 9,900.00$ -1% 10,000.00$ Flat
End of day 1 1.29250 5,575.00$ -43,69% 10,000.00$ Flat
End of day 2 1.28815 3,400.00$ -39.01% 10,000.00$ Flat
Trade close 1.30250 10,575.00$ 211.03%(5.75% compounded over 3 days) 10,575.00$ 5.75%

Let’s get this straight: both methodologies are “correct”.

BUT: Darwinex performance contains more information than Service Alfa’s calculation – the unrealised P&L that drives changes to liquidation value.

This has implications that matter.

Implication 1: Performance “masquerades”

Investors are always keen for brilliant performance. As always in life, there’s two ways about it: the proper way, and a short cut.

The safe one is to develop trading skills, manage risk with discipline, and work 2,000 hours a year to develop and maintain a unique strategy that delivers risk-adjusted returns. But hey, we know that’s hard work …

So why not take the short cut? If performance equals realised P&L and unrealised P&L doesn’t matter, why not “short cut” mind-boggling balance growth as follows?

  • Trade winning? Close it and credit realized P&L
  • Trade losing? Keep “corpses” in the “unrealised P&L” closet
  • Repeat

Of course, too many losing trades will trigger a margin-call (the worse the strategy, the sooner). Meanwhile, high return, low risk growth in balance will have lured plenty of profitable investors using service Alfa…

Implication 2: Risk management

A more fundamental implication is risk measurement. Measuring risk on volatility in balance leads A to these conclusions:

  • Volatility: low – A’s money “only” moves from 100% to 105.75%
  • Drawdown: A “experienced” no drawdown in this trade
  • Rally: A “experienced” a rally of 5.75% of his money on this trade

I.e. A will completely ignore that his monies moved around 40% over the 2 days of the trade, and that at the end of day 2 he was 6,600 USD worse-off.

Of course, Ace could argue that the trade worked out just grand, with a profit of 5.75%. But what if A was leaving for a Himalaya trekking trip and asked Ace to close any open positions at the close of day 2? Boy that would have been ugly!

Summary: why?

Calculating net liquidation values requires us to source & maintain the mark-to-market price database for all the assets you ever traded, and that’s not a small & cheap database to populate, host and maintain.

Further, the way performance is calculated on the BASIC and the DARWINIA statistics is slightly different, as Darwinia is a lot more analytically involved. You will occasionally see a few decimal differences between one and the other, as they are updated with different precision and time frames… but they’re not wrong.

But then again, that’s the only way to create D-Leverage, no DarwinIA, no DARWINs, etc. If it requires us doing things differently from service Alfa – so be it!


[1] For simplicity, we’ve ignored swaps & any other overnight charges.