The image of 'thumbs down' in the modern imagination, as well as much of what people (not least, by his own admission, Ridley Scott) think of when they think of 'gladiators', is primarily the product of Jean-Léon Gérôme's 1872 painting, 'Pollice Verso' (we'll come back to that name in a bit). Modern perceptions, including those in a number of history books, can give the impression that everything in gladiator fighting was very fixed and formulaic - that there was a precise 'running order' to how shows were organised, who fought whom with what equipment, and so on - and that there is essentially a single template of 'the gladiator show'. As with most things in the Ancient World, that simply isn't true - things were a lot more flexible and much more was done 'on the hoof' than is sometimes realised.

Firstly, to the famous signal. The phrase that comes down in much of the literature, pollice verso, (originally verso pollice in Juvenal's third Satire, written around AD 100-127) means 'with turned thumb' - Juvenal refers to how the crowd signalled their wish that a gladiator be killed, but not exactly how the thumb was turned. In his 1997 article, Anthony Corbeill collects citations for scholars arguing that it was turned up, turned down, hidden in the hand, pointed at the chest, or squeezed between two fingers [1]. His suggestion - which is the closest thing to a mainstream view on the subject - draws on the general association of 'thumbs up' in Roman culture with an obscene gesture called infesto pollice (the thumb being understood to represent an erect penis), and the use of the same term in certain texts to name the gesture used to call for a gladiator's death. So he believes that to signal death, you point your thumb to the sky (others have argued that this represents a drawn sword, ready to kill, versus a 'thumb down' signalling the sword ready to be put back in its sheath), and that the modern idea of 'thumb up for mercy' comes back to the unrelated 20th-century American use of that gesture. Instead, he suggests that you signalled mercy by pressing your thumb down against the closed fist.

A gladiator signalled his surrender by holding his index finger up, either to his opponent or to the editor, the aristocrat who had organised and paid for the games. Ultimately, it was he who made the decision, not the crowd, though our textual sources are unanimous that they made their opinions felt pollice verso as well as through more conventional means like shouting, jeering and waving whatever they had to hand.

It wasn't a given that anybody would win - we one account of an identifiable gladiator fight from the ancient world in a poem by Martial, in which he describes how two gladiators (Verus and Priscus) fought a long match, both acquitting themselves very impressively, and eventually both 'surrendered' to each other at the same time, whereupon the emperor Titus sent down two palms to indicate that they had both won. Another phrase that comes down through the sources is stantes missi - 'sent away, still standing' - and we see it faintly at the top of the depiction of a gladiator match on the Nimes Medallion, in a way that suggests it is being shouted by the crowd. We know that ordinary Romans had their favourites - several pieces of Pompeiian graffiti, as well as the gladiatorial epitaphs collected in this Oxford Reference article by Anthony Corbeill, make that clear; many gladiators' tombstones were paid for by their fans, or amatores.

Some fights were organised sine missio, but scholars are divided as to whether this meant that all losers were killed, or simply that all fights had to end in a clear conclusion and that both being stantes missi, with no winner declared, was not an option. At any rate, this was rare, and banned under the emperor Augustus, though we have sources that suggest it happened in later periods.

It's certainly the case that not every fight ended in death - mosaics and graffiti of gladiator fights are common, and often place a θ for the Greek word θάνατος ('death') next to defeated fighters - this would be unnecessary if defeat automatically meant death. Indeed, from the point of view of an editor, killing gladiators was costly - he usually didn't own them outright, but hired them from a lanista, the master of a gladiator 'school', and owed him far more if he returned his gladiators dead. In the second century, we know that gladiators who could be hired for 80 HS (small change for the sort of person who might put on a show; an ordinary legionary took home 3.3 HS a day) would require a payment of 4,000 HS (or 500 times more expensive) if they were killed or maimed. In some cases, a minimum price was set that no gladiator would be shown who fought for less than 1,000 HS - and given that dozens of fights could take place in even relatively modest shows, it's not difficult to see how bloodlust could quickly become ruinously expensive. In a really big set of ludi, such as those held in the Colosseum over the second half of AD 108, thousands could fight - an inscription on that occasion recalls '4941 pairs-and-a-half of gladiators'. As Donald Kyle puts it, 'gladiators were an investment, skilled artisans to be rewarded and not to be wasted.' [2]

Estimating the proportion of gladiators killed in the arena, or a gladiator's chance of surviving a given fight, is more tricky (in all fields, the evidence available from the ancient world doesn't usually lend itself well to precise quantitative answers). We can look at gladiators' epitaphs, for instance, in the knowledge that they are a self-selecting sample - many talk about dozens of fights, and one recorded as CIL X 7297 reads as follows:

Flamma, secutor. He lived 30 years. He fought 34 times, won 21 times, drew (stans) 9 times, and was spared (missus) 4 times. Syrian by birth. Delicatus, his comrade-at-arms (coarmio), made (this tomb) for a worthy man.

We don't know (though may perhaps guess from his young age) that Flamma died fighting, but even so, he came through 13 out of his 34 fights neither the winner nor a corpse. Estimates by Ville and Claval-Leveque, quoted by Kyle, suggest that between 15% and 20% of fights in the first century AD ended in death. However, Ville also considers that fights became more lethal over time, in response to greater audience familiarity with ever-grander spectacles and the consequent competition between editores to outdo each other with blood, gore and the expenditure that this represented, and that by the third century the death rate had increased to around 50%. Nevertheless, we know of many gladiators who survived to freedom and retirement, some after a huge number of fights - over 100, on occasion.