KEY WORDS: MuayThai, judging, nationalistic bias.

Data were collected over a one-week period from scorecards from the 2003 International Federation of MuayThai Amateur (IFMA) World Championships in Almaty, Kazakhstan. The competition was held in the Culture and Sports Palace from September 1st until September 9^th 2003. The competition included 44 countries competing for 54 medals across 14 weight classes. The competition comprised of A-class and B-class competitions; A-class featuring the world's top competitors and the B-class for less experienced competitors from countries with less experience in the sport. The data for this study comprised of tournament results from 70 A-class MuayThai bouts, each judged by between five and nine judges. The bouts featured 62 competitors from 19 countries and 25 judges from 16 countries.

Statistical analysis
We examined nationalistic bias at the level of individual judges scores for each bout.. As a given bout could have non-neutral judges sharing nationality with either boxer, judges were categorised as red (sharing nationality with the red corner boxer), blue (sharing nationality with the blue corner boxer) or neutral (sharing nationality with neither boxer). Scores for each judge in each bout were summed for each of the four rounds, where minus values were assigned to scores in favour of the boxer competing out of the blue corner and positive values assigned to scores in favour of boxers competing out of the red corner boxer. For example, if a judge had a boxer from the red corner winning three rounds and losing one, this would result in a score of +2. Conversely, if a judge scored a boxer fighting out of the blue corner all four rounds, this would result in a score of -4. In total, the dataset was made up of 2,028 difference scores over the 70 bouts. Of these 70 bouts, only the 43 with at least one judge sharing nationality with at least one of the competitors were used in this analysis.

These data were analysed using a multilevel model, with judges scores fitted as a normal response variable. Further details of this type of model can be found in Goldstein, 2003. In the current study, we fit a simple two level model with scores nested within bouts, and bout included as a random effect. Importantly, this controls for what are likely to be highly variable differences between boxer's abilities, acknowledging that judges scores are more likely to be similar within than between bouts. This type of approach is common where observations are clustered within groups (for example pupils within schools, or people within households) and these observations are likely to be affected by these clusters, which is certainly likely to be the case in the current study. Accounting for such clustering avoids tests that are often too liberal for level-2 (bout level) covariates and typically result in falsely rejecting the null hypothesis too often (Gibbons & Hedeker, 1997). The model had a single categorical predictor with three categories 'red' (judge shared nationality with the red corner boxer), 'blue' (judge shared nationality with the blue corner boxer) or 'neutral' (judged shared nationality with neither boxer).

Second, we examined the impact of any observed bias on the overall outcome of all bouts in the Almaty tournament and discuss control of nationalistic bias.

It is hypothesised that nationalistic bias is observed, although the overall is likely to be dictated by the ratio of the neutral to same nationality judges.

Nationalistic bias - judges scores within bouts
Of 70 bouts at the Almaty tournament, 43 (61.4%) had at least one non-neutral judge. Table 1 shows output from the multilevel model of judge's scores on the basis of whether or not they shared nationality with one of the boxers. Note, that negative scores indicate judging in favour of boxers competing out of the blue corner and positive scores indicate judging in favour of boxers competing out of the red corner (see analysis section). The model also includes a random bout parameter, again, as discussed in the 'analysis' section.

Firstly, the intercept value of -0.30 suggested that on average, judges sharing nationality with blue corner boxers (our reference category) scored these boxers around a third of a round better than boxers competing out of the red corner on average over the course of a bout. Of more interest though, is how scores changed with judge's nationality. Secondly, neutral judges typically scored boxers competing out of the red corner two- thirds of a round (0.64) better than judges sharing nationality with boxers competing out of the blue corner over the course of a bout. Thirdly, judges sharing nationality with boxers competing out of the red corner typically scored such boxers over a round better than judges sharing nationality with boxers competing out of the blue corner over the course of a bout, a statistically significant difference. The equation below provides a simple summary of the model;

Judges score = -0.30 + 0.64 'neutral' + 1.09 'red'

So a judge who shared nationality with a boxer competing out of the blue corner typically scored 0.3 rounds in favour of those boxers, a 'neutral' judge, who didn't share nationality with either boxer typically scored 0.34 rounds in favour of boxers competing from the red corner (suggesting boxers competing from the red corner were marginally superior overall) and a judge sharing nationality with the boxer in the red corner typically scored 0.74 rounds in favour of boxers competing out of that corner.

Nationalistic bias and bout outcome
Despite evidence of nationalistic bias, this does not guarantee that that the outcome of bouts will change. A simple, logical solution to the issue would be to simply remove non-neutral judges. This also allows examination of the impact of bias at bout level. Table 2 summarises the outcome of the 43 bouts with at least one non-neutral judge; using all scores for bouts but removing the scores awarded by judges who shared nationality with boxers competing out of the red or blue corners and using only neutral judges.

Table 2 shows only modest changes in outcome when removing non-neutral judges. Effectively, two bouts would change in outcome; one moving from a draw to the red corner boxer winning and one blue corner win moving to a red corner win. Essentially, non-neutral judges decisions impacted on the outcome of two of forty-three bouts where they were present, or two of seventy bouts in the entire Almaty tournament.

However, that is not to say that nationalistic bias does not have the potential to impact more severely on outcome. Table 3 examines the seventeen bouts with both red and blue judges and shows large differences.

The fact that the scores of judges who share nationality with boxers competing from the red and blue corners do not impact on the outcome is not a case of their scores balancing each other out (removing judges who share nationality with either red or blue boxers still results in 7 blue and 10 red corner wins), but is simply a consequence of same nationality judges not having enough judges (essentially a majority would be required if neutral judges are fairly consistent) for their scores to have real impact. Bias is clearly present, though same nationality judges' minority status reduces the impact of this bias. Moreover the natural control exerted by relatively large numbers of judges in the Almaty tournament also reduces the impact of unusual neutral judge scores, which could in turn potentially allow same nationality judges to have a greater influence.

The aim of this study was to explore national bias in MuayThai. The results suggest there was evidence of nationalistic bias in judge's scores at the 2003 World Championships. These findings are consistent, to varying degrees, with the findings of other subjectively judged sports (Ansorge and Scheer, 1988; Campbell and Galbraith, 1996; Popovic, 2000; Zitzewitz, 2002). The bias observed in MuayThai equated to approximately one round difference between opposing judges (i.e. one sharing nationality with one boxer and one with the other) over the course of a bout.

However, although there was evidence of nationalistic bias at the championships, the impact of this bias on the outcome of bouts appeared diluted by the large numbers of judges (i.e. there were few bouts where nationalistic bias had an impact on the final result). Although five judges are placed around the ring at all international IFMA championships, judges' scorecards are vetted by a jury panel that sits together and judges the fights (IFMA, und). This means in practise that each fight is judged by between five and nine judges in international competitions. In essence, judges sharing nationality with boxers at the Almaty tournament were effectively outnumbered by neutral judges, meaning that in real terms, the overall outcome of bouts was rarely influenced.

Although the impact of nationalistic bias on bout outcomes was diluted, the outcomes of five of the bouts were decided on by the verdict of a single judge. This allows the possibility that nationalistic bias could have played a role in the outcome of these five bouts. However, this only occurred in bouts that were judged by fewer than eight judges. This suggests that the current system can lessen the impact of nationalistic bias as long as eight or more judges from different nationalities are used to judge each bout. Moreover, many countries taking part in the championships did not send judges because of cost; this may have had the effect of magnifying bias associated with the host country.

To avoid problems of nationalistic bias, MuayThai may need to consider adopting some kind of electronic scoring system. However, the method used in amateur boxing would not be practical for MuayThai. Problems with judges failing to record all blows delivered in bouts, would only be compounded in MuayThai where judges have to record kicks, knees and elbows along with punches. On the other hand, a variation of the system employed in Taekwondo could prove to be suitable. Although it would not address all of the scoring issues in MuayThai, the use of body armour to record kicks and knees to the body electronically and judges recording strikes and kicks to the head and legs would be worth exploring. By using this system and reversing the points awarded in Taekwondo so body kicks and knees were worth two points and head (and leg strikes) one point, the system would enable amateur scoring to have some similarities with scoring in the professional MuayThai. If this system were to be adopted, different impact tolerances would need to be set for the different weight classes. However, this would not necessarily address the problem of recording scores for techniques that currently score highly like unbalancing an opponent with a front push kick, that involves good timing rather substantial.

Future research will investigate the impact of other ingroup biases. The potential for nationalistic bias to be compounded by other ingroup biases is real. In this study 46 (65.7%) bouts were judged by at least one judge from a neighbouring country. Where biased judges have the opportunity to outnumber neutral officials the outcome could be seriously biased.

Our study shows that nationalistic bias is present in international MuayThai judging. However, the number of neutral judges currently used to judge bouts at world championship level means that this level of bias generally does not influence the outcome of bouts. Our results suggest that the officials responsible for assigning judges to judging panels need to consider the nationality profile of panel members for each bout carefully.

• Nationalistic bias is evident in international amateur MuayThai judging.
• The impact on the outcome of bouts is limited.
• The practice of using a large number of neutral judges appears to reduce the impact of nationalistic bias.