Table of Contents Abstract Introduction Background The Sandbox The Data A: Transaction Graph B: Bow-tie Model C: The Sandbox Support D: The Sandbox Scandals E: Whales Methodology Results Related Work Conclusion Resource Contributions Acknowledgements and References Abstract Abstract Abstract Introduction Introduction Introduction Background The Sandbox Background Background The Sandbox The Sandbox The Sandbox The Data A: Transaction Graph B: Bow-tie Model C: The Sandbox Support D: The Sandbox Scandals E: Whales The Data The Data A: Transaction Graph B: Bow-tie Model C: The Sandbox Support D: The Sandbox Scandals E: Whales A: Transaction Graph B: Bow-tie Model C: The Sandbox Support D: The Sandbox Scandals E: Whales Methodology Methodology Methodology Results Results Results Related Work Related Work Related Work Conclusion Resource Contributions Acknowledgements and References Conclusion Conclusion Resource Contributions Acknowledgements and References Resource Contributions Resource Contributions Acknowledgements and References Acknowledgements and References Results Results A. Bow-tie Model A. Bow-tie Model Our complete network has a total of 659, 248 nodes. When we split the network into the components of the bow-tie model, we find that most of the network falls into the SCC and OUT partitions. The SCC accounts for 390, 531 nodes (∼ 59.25%) while the OUT accounts for 264, 501 nodes (∼ 40.13%). The IN only accounts for 4, 071 nodes (∼ 0.62%). The IN TENDRILS, OUT TENDRILS, and OTHER groups account for 7, 136, and 2 nodes respectively. We found the TUBES partition to be empty. Since most user transactions in The Sandbox involve some transaction in response, we don’t see many addresses in the IN category. It mostly consists of inactive or bot addresses and addresses approving the use of the SAND token for the tradein of another contract. On the other hand, the OUT group holds more investors than normal users. One would have to interact with The Sandbox to get anything from its game aspect, so this makes sense. Then, in the SCC, we have the bulk of it making up The Sandbox’s gaming community. These primary users drive the value of SAND by actually using it to play in The Sandbox B. The Sandbox Support B. The Sandbox Support After calculating the number of transactions per day for the entire range of our dataset, we plotted the values for the 30 days before and after the dates attributed to each “support event” in figure 2. We did the same for the total value of the transactions each day in figure 3. Both graphs show short-lived spikes in transactions and value but no long-lasting effects after these spikes stabilize. The two exceptions to this are deadmau5 and Warner Music Group. However, the consistent increase in daily interactions after deadmau5’s support is likely attributed to the simultaneous GameFi boom. On the other hand, Warner Music Group’s effect is not so easily written off. Nevertheless, the rest of the support events quickly returned to the norm, meaning they failed to attract a consistent user base. To take a closer look at these results, we have graphed the data for deadmau5, Warner Music Group, and Square Enix in figures 4, 5, and 6, respectively. Here, we have split the transactions based on whether they belong to the bow-tie model’s SCC, OUT, or IN groups. In the deadmau5 graph, figure 4, there is an initial spike in SCC transactions followed by a dip that stays well above the number of transactions each day throughout the previous month. This indicates a reliable increase in activity. There has also been a noticeable increase in the number of transactions belonging to the OUT group. It is not as significant as the SCC’s but still significant relative to its previous values. This indicates an increase in investors. The IN group does not have a noticeable change in activity, which could be attributed to its small size. However, we do have to keep in mind that this support event coincided with the widespread boom of the GameFi market. There is a similar pattern in the Warner Music Group’s graph, figure 5. There is an increase in activity in the SCC and the OUT groups without anything noticeable in the IN group. However, the increase in OUT transactions is less noticeable since the previous month’s values are higher than they were with deadmau5. Regardless, this again indicates an increase in the active user base and investors. Additionally, this effect cannot be as easily attributed to another event, so it is likely due to the Warner Music Group’s efforts in The Sandbox. We use Square Enix’s support to represent the rest of the support events, as they all had similarly unspectacular results. In the Square Enix graph, figure 6, there is a very short-lived spike in SCC transactions followed by an immediate return to normalcy. The OUT and IN transactions are not noticeably affected. The other support events saw similar results6 except for The Smurfs, which had no effect, likely due to how early into The Sandbox’s life they supported it. Therefore, Warner Music Group is the exception rather than the rule, and most of these events did not have a long-lasting effect on the activity in The Sandbox. C. The Sandbox Scandals C. The Sandbox Scandals We created similar graphs of the number of transactions and the total value of transactions for our scandals in figures 7 and 8. In both graphs, we do not see any long-lasting effects due to the CEO’s Twitter account being hacked or the Employee’s computer being hacked and used to send phishing emails. In the value graph, the same applies to the SEC naming SAND unregistered security and the Ronin Hack. However, the effects of these scandals are more noticeable in the transactions graph, with the Ronin Hack having long-term effects. Similarly to what we did for the support events, we have graphed the data for the Ronin Hack, SEC naming SAND a security, and the CEO’s Twitter account being hacked in figures 9, 10, and 11, respectively. Once again, we split the transactions based on whether they belong to the bow-tie model’s SCC, OUT, or IN groups. In the Ronin Hack graph, figure 9, there is a delayed spike in SCC and OUT transactions followed by a dip that drops slightly below the previous days. The reason for the delay is that the Ronin hack took about a week to be noticed. This result indicates that the initial spike was due to users and investors jumping ship since there are fewer daily transactions after the spike. The long-lasting effect of this scandal is clear despite The Sandbox not being directly affected by the Ronin Hack. In the SEC graph, figure 10, there is a massive spike in SCC transactions a few days after the event, but there is not much of a change in the number of daily transactions afterward. This may seem odd, but it is worth noting that this event happened far after The Sandbox’s prime, so its user base was barely active. This is further evidenced by the lack of a reaction in the OUT group, which indicates that most of the investors in the OUT group who would have been scared off by this event had already left. We use the CEO’s Twitter hacking as a representative for itself and the Employee computer hacking since they had similar results7 . In figure 11, the hack has no noticeable effect on the activity in The Sandbox. Like with the SEC event, it is important to consider that this scandal occurred well after the prime of The Sandbox, so the user base was already at its weakest point, with the users left not being shaken by these sorts of scandals.  D. Change in Bow-Tie Model Over Time In figure 12, we graph the changes in the categories of the bow-tie model over time in our Sandbox network. We also include future addresses not yet involved in the network to illustrate when new addresses will be introduced. Note that addresses cannot leave the SCC once they have entered it. The groups other than the SCC and OUT are small throughout the graph. The SCC grows steadily between events, but we see three significant jumps after Adidas, Steve Aoki, and the SEC hearing. The Adidas jump can likely be attributed to the GameFi boom. The Steve Aoki jump from the OUT group could, in part, be attributed to the lasting effects of the Ronin hack, causing members of the OUT group to sell their investments to members of the SCC. In contrast, the new addresses that entered the SCC could be attributed to new users taking advantage of the dropping value of SAND. The influx of addresses from the OUT to the SCC after the SEC hearing is likely a result of investors selling their investments, often to members of the SCC. The growth of the OUT is less consistent than that of the SCC, partly because the OUT can lose members while the SCC cannot. Nevertheless, there are several significant increases in new OUT addresses. The first is after Adidas, which, like with the SCC, can be attributed to the GameFi boom and, thus, a significant increase in investors. We see a decent increase in new OUT addresses after the Warner Music Group support, which matches what we saw in our closer analysis of this event in figure 5, where we saw a slight increase in OUT activity. We see a similar pattern between the Ronin Hack and Steve Aoki as we did with the SCC, which once again indicates that investors may have been hoping the effect of the Ronin Hack was just a dip that would be recovered from8 . After the Gucci Vault event, there is a surprisingly significant increase in the OUT group. Although this does match a spike in OUT transactions before the Gucci Vault event present in the data9 , this cannot be attributed to the Gucci Vault event since it occurred afterward. There is also another surprising increase in the OUT between the Care Bears event and the Employee’s computer being hacked. This is likely due to a slight increase in value that SAND saw between these two events10 which attracted investors again. It is difficult to attribute this increase to either of these two events since they did not cause any significant increases in OUT group activity11 . Overall, the SCC’s growth is more consistent than the surprisingly unpredictable OUT. It is generally difficult to attribute the significant increases in addresses to the events explored in this paper since the increases generally do not match the increases in activity that resulted from these events except for the Warner Music Group support. E. Whales We found 2,464 whales invested in The Sandbox as of October 26, 2023. Of these whales, 890 are in the SCC, and 1,574 are in the OUT group. However, we also check to see if any addresses ever met this requirement at some point during the three years our data covers. There were 33,758 qualifying whales over time, with 32,174 in the SCC and 1,584 in the OUT group. As we can see, most of the former whales were in the SCC, which makes sense since the SCC would be the most likely group of nodes to offload one’s SAND on. We combined the transaction and value graphs for the support events and scandals in figures 13 and 14, respectively. We combined the graphs because they show almost identical patterns to the normal graphs, so the same analysis applies. This implies that whales are largely involved in most transactions in The Sandbox, which indicates that some of these whales could function as resources for The Sandbox. To avoid filling this paper with more graphs, we can see how these patterns will also hold when we include the past whales since they will just be adding more transactions to a pattern that already matches the pattern created by the entire network12. This is further supported by the average degree of the network compared to the average degree of the whales. The average degree of the network is about 12.866, 5.395 when we exclude all whales, while the average degree of whales is 735.520, 151.286 if we include past whales. This demonstrates that whales interact with significantly more addresses than normal users. Authors: Fernando Spade Oshani Seneviratne Authors: Authors: Fernando Spade Oshani Seneviratne Fernando Spade Fernando Spade Oshani Seneviratne Oshani Seneviratne This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. available on arxiv available on arxiv
